Back to Community
How do we explain beating the S&P500 by a factor of 30, having the highest returns in the contest in conjunction with the highest Sharpe, and rank 320th in the contest? This is after less than one month.

I am seriously struggling now to understand the below. Can anyone help me understand how this is even possible?

This is me -
This is the leader -

The leader has an annual return of 0.0001324%. You can get more money if you buy a government bond. What in the world is going on here? I have made 3.3 milllion dollars on 10m in less than one month. There is no way that this can be correct. Are we evaluating the number of zeros between decimal points? Can we bring living trading back please? This entire endeavor is starting to seem pointless. At least then we can profit from our own ideas. If I have the highest Sharpe, and Sortino ratio, based on the definition of both, how can this contest be seen as a legitimate assessment of how good an algorithm is, given this scenario? How in the world can the Beta of an algorithm be used to determine it's overall value? If an algorithm doesn't even beat the benchmark, how can it possibly lead in a contest?

Sharpe Ratio

The Sharpe ratio uses standard deviation to measure a fund's risk-adjusted returns. The higher a fund's Sharpe ratio, the better a fund's returns have been relative to the risk it has taken on. Because it uses standard deviation, the Sharpe ratio can be used to compare risk-adjusted returns across all fund categories.

Sortino Ratio

Just like the Sharpe ratio, a higher Sortino ratio is better. When looking at two similar investments, a rational investor would prefer the one with the higher Sortino ratio because it means that the investment is earning more return per unit of bad risk that it takes on.

49 responses

Aniken's points are all valid. And, as Vladimir is stating: “until they change the rules, it will be like it is and like it was”. But, that is the whole point. To let Q know that with the current rules, the present contest leader is making a whopping $37.17 on a $10M dollar investment.

I would add, and categorically state, that even at a 6x leverage they won't make any money on this one. No client of theirs would ever accept such a strategy whatever they would like to stress as redeeming qualities, if any.

It is ridiculous to see the leader generating $37 bucks and winning the contest while someone else is generating $3.3M, under the same testing conditions, relegated to the 77th spot.

Something is wrong. Something needs to be fixed. Something needs to be justified.

One could view this as volatility anxiety or outright fear of single digit drawdowns. But, whatever, Q is missing the boat if the present contest leader turns out to be the ultimate winner having similar numbers as presently shown.

If I were in 77th place, I would probably not have said anything. I'm currently ranked 320. Sharpe and Sortino measure risk/reward. The current leader has a Beta rank of no. 1, which clearly indicates that the weight of the overall scoring is grotesquely skewed and has completely missed the boat. Higher returns require larger risks. That is a mathematical fact. The question should be, is the risk taken worth the reward. With a Sharpe of 10, currently no. 1, and a Sortino ranked 2nd, the Algorithm is screaming, "No Justice. No peace!" No human being in their right mind would look at 3.3 million in one month and take 37 dollars. The system is beyond broken if this plays out as it currently stands. Also the lack of transparency is also concerning. I do not have a draw down of 26 percent in my algorithm. So, I don't even know where that came from. They say this is live trading, but I can see the stats from live trading right in the console. See below. Something isn't adding up. Not only do I not see this in live trading, which I thought is what paper trading was, I can't make any sense of what this contest is attempting to score. At this point, its extremely discouraging, because it would appear that this is a Beta Contest, rather than what is should be about, Alpha. I'm starting to believe that this platform has lost its way. And I'm not the only one.

Again guys, I'm not trying to slam Quantopian. I think it's a fabulous idea. I wrote hundreds of algorithms, before I found patterns that I could reliably exploit for my own trading. However, I don't think that any serious developer would waste his/her time with inconsistencies like this. Also, this is the sort of thing that could lead someone to believe that this entire thing is simply a scam, set up to steal good algorithms, while rewarding bad ones as a diversionary tactic. If this contest ends and I double the investment, which is highly like at the rate my algorithm is progressing, what conclusion can a rational person draw? I debated for a month before entering the contest, and now I regret that I ever did. The markets have been soaring. Obviously, if your algorithm soars with it, should we get punished for a high beta for capitalizing, and ignore other critical statistics? What about the back tests, in which I was able to make 207% through the market crash of 2007. I was honestly hoping for an allocation, but if $37 is what we're looking for, I certainly wouldn't be looking forward to a $3 allocation return. And the worst part about all this, is that your scoring system is actually encouraging developers to code this way, which is counter conducive to investor ideals, particularly those who are looking to beat the market. I said it before, and I'll say it again. If your algorithm doesn't beat the SPY benchmark, it's unsellable. And if the point of this platform is to obtain algorithms to sell to investors, and the contest is a grading system, how does any of this make sense? How could a rational person conclude that this platform is credible at this juncture?

This contest is for long short equity strategy algorithms.

An algorithm that is not hedged might not do well in the contest.
"Your algorithm must be hedged to the market."

Dan already pointed this to you in one of your earlier posts on the subject.

You left the other piece off, "or be entirely in cash." My algorithm does not borrow money.

This contest is geared towards a long short equity hedge fund that Q is creating.

Read the allocation page for the list of criteria that an algorithm need to satisfy
"As a practical matter this means that we're looking for algorithms that have a strong hedging component, or are completely hedged at all times."

100% long exposure is an assumption. I assure you that assumption is false.

Check the images you posted in the other threads. One of them states X% long, 0% short.

Again, you are making a lot of assumptions. What you deem impossible is also an assumption. No mans intellect is bound by assumption, unless he subjects himself to the perception of such boundaries. If I were to divulge to you how I was able to achieve what I did, I assure you that you would have no problem understanding. Unfortunately, I am curious to see how this plays out. There are various ways to protect yourself from losses, and I assure you I am doing precisely that. Let's imagine that I am able to double, or triple 10 million, without exceeding the thresholds for Max DD, while maintaining the highest returns, Sharpe Ratio, and Sortino, with a spectacular win ratio on all trades. Would you deem an algorithm that is completely hedged with a return of less than 5%, that fails to beat the SPY more appealing to market to your clients, because it takes less risk, while failing to outrun inflation? The rational question I believe should not be how much risk is taken, but rather, is the risk taken worth the reward. If Quantopian either isn't willing or does not have the capacity to market an algorithm that outperforms the market by a minimum factor of 10 in a flat market, and factors far beyond in a bull market, it simply isn't trying to. To not present something like this as an option to a client implicates one of two things. Quantopian owners are either flawed in their conceptual thinking of assessment, or worse, hold nefarious intentions to the users of their platform. Not a single investor in their right mind comparing two investments neglects to examine the annual return or Sharpe ratio and looks solely at Beta. My name is also Leo by the way.

Again, you are making some assumptions. The algorithm isn't a "long" only algorithm. However, I have a hard time believing that Quantopian has zero interest. My theory is that they fully know and understand value. If they don't, no harm, no foul. At that point, I may simply divulge the algorithm to the public, rendering the strategy to every market maker in existence, and building a site for all to see. Imagine the sort of traffic that would draw. I'm also curious to see how quickly the market would adapt. It may render it useless, in which case if Quantopian did have nefarious intentions, those would immediately be neutralized. I think I can predict with some certainly how to capitalize on that as well. Either way, it's a win win situation for me.

@Leo Williams,


Quantopian open contest 26 results

Best Return

Taupe Eagle
Yusup Jorayev
score_contest_rank 1025 score_contest 21.23
Paper Trading
score_pt_rank 1014 score_pt 21.23

rankAnnRet_pt 1 annRet_pt 152979.79%

Just imagine that number $15,297,979,000
it is 5000 times more than yours but he was ranked 1025.

I think you're missing what I'm stating.

The algorithm you selected has the following:

  1. Max DD of -9.39291707. 939%. literally lost 9 times the money.
  2. Sharpe: 0.054763485 - This clearly states the risk is not worth the reward

My algorithm has the following:

  1. Max DD of 0.26 - This is supposedly in paper trading, which I do not see on my screen. There is no such loss, but even so, it is still within the constraints.
  2. Sharpe of 10 - Number 1 ranked Sharpe. One is generally considered good. Two is terrific. Ten was inspired by God himself.
  3. Ranked 11th in Stability - 0.9051
  4. Ranked 2nd in Sortino
  5. Ranked No. 1 in Annual Returns
  6. 32.70% Returns $3,270,221.65 Dollar P/L from Oct 19th - Nov 17th

Do you see the difference? Or are you making the case that these algorithms are equivalent?

This is me -
The other guy -

Its very clear that Quantopian wants specific criteria, you do realize that you can take your algo and go run it yourself? I have never even bothered with Q's contests since its clear they value long/short over total return, even if you "hedge" by pulling out of the market based on signals. It is what it is, if you really think you have something valuable, go take it elsewhere, but getting mad that the rules don't favor your style of algorithm is just wasting your time.

If you can get consistent 30% monthly returns, start trading! For what its worth, Q was just in the news as their fund is actually down 3% on the year, and there were some shakeups in the org as a result.

Thanks Kevin. That doesn't surprise me at all. I tried to research to see if I could find anyone who had gotten an allocation, and the only thing I found was Jedi Enforcer Dan Dunn promoting the platform on Quora, and other people stating they did not trust Quantopian. As a result, I have a hard time believing anything coming from Quantopian at this point. The idea that this has no value, and to try and convince logical people of the same, just makes them look utterly deceitful and ridiculous. None of the thousands of people who have read this, and other related threads on my Linkedin, or elsewhere buy this, and everyone is telling me that in their personal opinion, the entire site is a scam. I had one guy who works in finance just bust out laughing. He said. Wow! At least now you know. I don't know what to believe, but I can tell you that I do not believe there is any rational explanation for what is happening. The Sonic the hedge hog analogy might make sense, if we had hedge hogs that moved faster than snails. Nevertheless, I have a small funds algorithm that did extremely well this month and I am trading my own money. 105% return on the month.

Also, I do disagree with you on the point that Quantopian "clearly" wants specific criteria, because the algorithm meets and exceeds all of their listed criteria. And as much as I would like to give people the benefit of the doubt, what is slowly emerging in my mind as an alternate hypothesis, based on the feedback of my colleagues, friends, and family is that this site is set up to steal from its base, by parading useless algorithms as forerunners, and dismissing useful ones as useless. The developer then goes away and trades their own money, while Quantopian capitalizes in secret. They then pay nothing to the developer and cash in on all their hard work. I believe it was King Solomon who said:

It is nought. It is nought said the buyer, and when he has gone his way, he boasts. As far as I'm concerned, this algorithm is the Deathstar, and if planet Quantopian does not surrender to reason, its entire reputation will be destroyed by virtue of its own proclamations. All risk averse Jedi with wedgies, huge bifocals, and quadruple decimal draw downs and returns will be seen as outcasts and lunatics! Embrace the dark side!

I guess that is one way to look at it. I found quantopian after leaving the algo/HFT world, and I started building out a platform to do trading on my own (since I was then no longer constrained from trading due to company or regulatory restrictions) based on top of Interactive Broker's APIs and pulling down data from yahoo finance and a few other free resources. It was a really slow process, and in the midst of googling something, I came across Quantopian, and immediately felt like an idiot- the data was way better packaged than what I had, they already had a connection to do paper and live trading with IB, and their backtesting platform would have taken me years to get to that point.

I personally always saw Q as a backtesting platform primarily for my personal use- and if an algo somehow caught the interest of the men behind the curtain and they wanted to give an allocation, great, but I didn't need their capital, and honestly, would rather keep the lion's share of the returns. Its unfortunate that you can no longer connect directly to IB and run your strategies, I now run two of mine more as a screen with manual rebalancing on a monthly basis.

My point I guess is- it sounds like the platform has allowed you to find a potentially very lucrative strategy to make money. You sound kind of angry at Q, but they have provided you tons of value, what is there to be mad at?

Based on your performance stats and assuming they will hold for future out-of-sample you could make one simple change to your algo that would potentially move you to the front of the pack. It's right in front of you. Can you find it?

Sharpe of 10...

Dude you are printing money, why even bother with Q!

That is a very compelling perspective, and I have embraced that aspect fully. What I find grotesquely offensive, are the perpetual insults to my intelligence, and not just me, to any developer reading these threads, with the naive aspiration of taking Quantopian up on their offer, in the quest for mutual benefit. As far as the Lion's share goes, that part of your analogy is false. I built this thing to scale into the hundreds of millions. I do not have hundreds of millions of dollars. However, Quantopian's clients do. So, if they take this behind the scenes to a ten million dollar client and sell it, while telling me that it's useless, they are the ones who are getting the Lion's share, and doing so in diabolical fashion. If that happened to you, I suppose it may trigger a mild emotional reaction, but there is a big difference between 1500% of $10,000 and 10% of the returns of a $10,000,000 allocation. And with a return of almost 70% per year, that would be 700,000 per client. Suffice it to say, I am mildly annoyed at this fiasco.

I know I could do that Harlequin. In fact, I already have, but Q is showing itself to be so ridiculous at this point, why even bother entering another contest. If I can't trust them to do the right thing now, why would I trust that they would do it in the future? I think the only recourse is to publish it and make it useless, and explain to the world what happened here. That should draw enough traffic to put a stop to this debauchery, and ensure no one benefits, garner me merit for my hard work, while I use my other algorithms to grow my account.

the algorithm meets and exceeds all of their listed criteria

Beta is .7730, significant weight in the score based on what the investors have evidently said they want.

The long term beta of the algorithm is .14. You even said it yourself that it would come down. Secondly, the markets are soaring. I will make no apologies for soaring with it. Lastly, I am an Alpha male. I make no apologies for that either.

Aniken you have three choices now:
1 - if you have total faith in your algo put all your money in it and in few years (compound return) you will buy Q, IB and whatever you want.
2 - Publish the algo and let the all world to judge your reasons and maybe to proclaim you the new MIDA king.
3 - Go out with your girl friend and forget Q (they are a private company, with expenses, and correctly they have full choice of the business)

I know what I will do: I will continue my English course and the algo-trading hobby :)))
ps: if you choose the 1 please let me know, I will invest on your algo


Darth Vader is right though.


I think that (if I’m reading this correctly) you have a 1.248% drawdown not a 124.8% drawdown. Which would be totally legitimate and within expectations from both parties (pretty damn good for a return of 1147% out of sample albeit only a month). Is this still the contest entry you’re referring to?

I don't know how they generate their numbers. And you're right, it does say 1.248%. That was my my apologies to the people a Q on that. However, there is no transparency as to how they arrive at those metrics. I am live trading the same. I can see it in the console. It says 33%. I suppose that is an annual projection, compounded based on current standing. Again, I do not know, but based on what I see, it's made 3.3 million on 10 in a month. We'll see how it shakes out.

Aniken Leo,
You are clearly one of the best traders in the world, with the capability of generating a higher return each month than lesser traders & investors such as George Soros & Warren Buffett can even come close to making in a year. At 33% compound rate of return per month, if you start now then in 5 years time you will multiply your money by a factor of 1.33^(12 mo/yr*5 yrs) = 26,983,510 times. Even with a starting account of only $20k now, your net worth in 5 years time will be more than half a trillion dollars. Another year or so after that you will have made enough to pay off the entire US National Debt, and then another year after that the sum total of all of the World National Debts combined, if you so wished to. But where to from there? If you are a generous person and you donate even a small percentage of your gains to charity, then the world's problems should all be solved within less than a decade.

That sounds a bit over the top Tony, even for a Jedi.

;-)) ..... well perhaps just a bit .... and of course it does depend on the scalability of your algo. But nevertheless within a relatively short space of time, you should be able to take control of the entire Northern Hemisphere. And as to the southern hemisphere, well there's not much there really, just a few desolate places like South Africa & Australia with some kangaroos hopping around and lots of big sharks at the beach.

Einstein theorized that light would be bent by space time, in accordance with the dark side force of gravity . It is now clearly understood that nothing can escape the event horizon of a singularity. The markets are no different. You can only acquire so much mass at a given point in space time. You can only transform so much mass at any given point. The liquidation of an asset parallels the sublimation of solid state matter. The larger the mass of the asset, the more energy required for liquidation. If we take the initial investment of even the most potent algorithm, we will see that the growth of the current sum, as it traverses through space time, irrespective of how initially parabolic, will endeavor to bend towards equilibrium. Therefore the arch of even the greatest algorithm will always bend towards a zero percent gain. You cannot scale beyond the even horizon of the market. What are the odds that a programmer so financially inept and grotesquely primitive, coding for only six months in python, would venture to incorporate logic into an algorithm to compensate for such an anomaly? I wonder if the universe is hedged somehow?

Aniken, indeed it is hedged, before you get too close to the event horizon, would you mind sharing what general strategies you use? We are in awe

I love this thread <3

Hi @Jens, yes, isn't he great. Our friend Aniken Leo keeps us all enthralled, wondering just what he will surprise us with next!

Hi & greetings, @ $kywalker Leo.
I noticed in a post of yours here about a week ago you stated: "I wrote hundreds of algorithms, before I found patterns that I could reliably exploit for my own trading. However, I don't think that any serious developer ...." Now, when I do a R-mouse on your name, Q shows me there are more than nine thousand backtests, but only 86 algos. Presumably you simply deleted all the ones that weren't good. However as part of the process of developing good algos always involves going down a lot of roads that are not sign-posted as being dead-ends, don't you think some of the bad ones could also have served as useful learning tools for others, at no cost or risk to you. Did Q enforce a quota or limit to the number of algos that you could actually retain on the system? [I think i seem to remember seeing or reading something vaguely similar with regard to @Vladimir also, although maybe i'm wrong about that]. Can you advise please.

Not sure if there is a Quota. My journey began in Ruby on Rails, where I wrote a moving average crossover tool to scan the market. However, the first algorithms I wrote were in Ninja Trader. Following that I started using the Yahoo Finance API and JavaScript to look for patterns. When yahoo discontinued it's finance API, I decided to take a serious look at Quantopian. As far as Vladimir is concerned, he sent me an email a couple weeks back stating that he was interested in purchasing the algorithm, said he had a big budget, and understood my frustrations with Quantopian. Who he is, what he does, or what his true agenda may be, I do not know. I just assume that everyone I hear from is from Quantopian. However, let's be frank. It's not my credibility here that is in question Tony. I'm simply just a developer who was naive enough to believe in Quantopian, thinking that if I wrote a good algorithm, it would receive an allocation. I actually believed and trusted these guys. And after 4 million in profit, in a single month on 10 milllion, I am being told that my algorithm is inferior to one that has made a whopping sum of $37. I don't have to say another word. Quantopian has destroyed its own credibility and if you honestly consider what has happened here, it is an emphatic disgrace.

I'm pretty sure Quantopian is into making money. But they are using their clients' money, with their very specific needs. Also, they state everywhere that they need at least 6 months of OOS data to properly evaluate an algo: they don't know you, they don't look at your code, so that's the only way they can confirm it might actually work as intended.

@Aniken $kywalker
You have great returns with less risk , I think you can include some diversification assets to reduce DD and increase sharpe.
If you not using leverage here , with this algo you can easily find allocation in other places . Don't bother with Q restrictions , they want bond like algos , with less DD .

What DD? 2% DD, 3.5 million in profit in 30 days on 10 million, No. 1 Sharpe, No. 1 Returns, No. 1 Stability. Seriously? Honestly, when I read that, my first thought is, "Are these guys from Quantopian too? Are they bots?" Why would someone invest in an algorithm that doesn't outperform inflation, or the S&P500, and is more risky than a government bond? They wouldn't period. So, why state that as a goal for the site? It literally screams, "I'M A BOLD FACE LIAR WHO CAN'T BE TRUSTED AND I WILL INSULT THE INTELLIGENCE OF EVERYONE ON HERE, BECAUSE I BELIEVE MY USER BASE TO BE EMPHATICALLY STUPID!" They also wouldn't choose $37 over 3.5 million. Even a person who is legally retarded could make a proper distinction in that case, but guess what? We're not all stupid. Some of us understand that if you will make a statement like that, that you yourself should know is absurd and dishonest, nothing else about this site can be considered credible. My personal belief at this point is the the people at Q are either out of their mind, or are simple the lowest form of human life. Either Quantopian is peddling a narrative that a 4 year old can deduce to be ridiculous, or it is literally a site set up by the biggest losers, in the history of the free world, to steal what they lack the intellect to create themselves. If the latter is the case, Q would essentially be a criminal enterprise. Nevertheless, we must give the benefit of the doubt and let things run their course to reveal the truth of the matter, before we pass sentence.

If you're ranked lowly, it may be because you don't meet all three requirements. Are you always hedged? Is your beta between -0.3 and 0.3? Does your algorithm make money in the backtest window they chose? If the answer to any of those is no, you haven't met the requirements for the Quantopian Fund/Contest and as a result shouldn't expect to do well.

When I checked your entry, although is is doing well, it isn't hedged at all times. You need to bear in mind that Quantopian is looking for something very specific, not just anything that does well in backtesting and out of sample live trading.

If it is really that good, use it yourself and leave the site. I don't see why you are still on here complaining if Quantopian is "peddling a narrative that a 4 year old can deduce to be ridiculous, or it is literally a site set up by the biggest losers, in the history of the free world, to steal what they lack the intellect to create themselves". At the moment your sole purpose on this site seems to be complaining about being on the site. Do us all a favour and log off.

This is what I wrote, currently sitting at 3.6 million.

Clearly, you didn't read the entire thread, though you implicated that you did. The answers to your questions were posted in the link above. You also seem to speak for a large audience, when you say "Us", but I'm not sure who that audience is. It's OK to express your opinion, as we are all entitled to that. However, you should always endeavor to speak for yourself, when casting your perspective on who should and shouldn't log in, and who favors would be done for. I will however, give you credit for the intelligent statement you made in the "use it yourself" suggestion. That is precisely what I am going to do. However, it is also import to the rest of "Us", who have, or plan to, develop on on this site to understand who, and what we are dealing with. That's called being considerate to your fellow developer. Now, I will pose a question to you, before I log off and leave you in euphoria. If you're "honestly" a Quant and a trader, what is your goal in using this site? Would you rejoice if you made $37 as a trader? Would wall street hire a Quant with returns in this ball park? Here is the person currently in first place:

This individual has a 0.0003591% YTD return, making $3,591 on $10,000,000. If we invested the same in the S&P500 index, we get 16%, and $1.6 million dollars. If my Quant did this for me in a year, he's be thrown from the roof, but you're right. Q is not the place for algorithms that make money. In fact, I just read a recent, very convenient, addition.

Low Exposure to Style Risk

We want algorithms that limit their exposure to well-known investment styles such as investing in small market capitalization companies over large companies. There are five investing styles we track: size, value, momentum, short-term reversal, and volatility. While there are valid strategies that exploit consistent exposures to these investment styles, we are looking for algorithms that limit their exposure as much as possible and seek to exploit novel relationships in the market to achieve consistent profits. As a rule of thumb we prefer algorithms that limit their maximum exposure to any single style to less than 20%, and ideally target an average exposure of as close to 0% as possible.

So, you cannot trade small caps. You cannot capitalize on the momentum, volatility, or reversal of Large Caps. So essentially, we need to eliminate the most profitable opportunities and seek strategies that are essentially stagnant. So, you are right. A 0.0003591% return is exactly what they are looking for. However, you can earn more money than that with a government bond, which is considered to be the most stable investment. So, I ask you to think young Padawan. Who sets up a company to sell something that no one in their right mind would buy? And if we can conclude that to be the case, is it wrong to question the true purpose of the platform? Quoting the great Bob Marley.

You can fool some people sometime, but you can't fool all the people all the time!

Logging off. :-)

@Aniken $kywalker

I think you misunderstood my statement. If you have good algo , then why not to negotiate with other Asset Management Companies.
Not only Q have funds , there are a lot of money in the market that to be managed properly.

Lol. Q has taught me a valuable lesson. Never trust anyone period! End of story. I will say this though, based on something I thought was a bit enlightening earlier in this thread. I now have the key to make millions of dollars with several algorithms I've composed and copy-written. Without the Quantopian platform, I would have never been able to achieve it. It really doesn't matter what Q does. It doesn't hurt me at all. If they truly are set up to steal, the word RISK would be truly comical, given what could happen with something like that. In a weird way, I sort of have to thank them. I've been running parallel tests this entire time, since I composed my algorithms. 187% on the first for small funds in a month and a half, and 3.5 million on the second, which is the subject of this thread. When you think about trading, the biggest dilemma you face is figuring out which way to go and what to purchase. I can pick that now with close to 85% accuracy. No sense in being angry at me. I achieved what I set out to do, by creating the impossible. I said it from the beginning. The market is nothing more than a function, with inputs and outputs, and the objective of of the algorithmic trader is to amplify that input, while minimizing risk. I plan to retire in a years time. And if I should wake up one day and Dynamo Dan and Spacely Sprocket Fawcett are being sentenced before a court of law, I'll tell them to keep a close eye on the Alpha, be sure to minimize their beta, watch very carefully for the DD, and feel the power of the dark side! Lmao!!!

All fun an joke aside though, Quantopian is an awesome site. You may never get an allocation, but the tools they give you are unbelievable, and what you can accomplish with them is incredible. I have to admit I've been sort of ungrateful. I've learned a lot here and it's going to change my life in a way that I never dreamed. I know I've been a pain in the .... but, I hope I've added some value in letting people see what's possible with a site like this. If I've inspired even one person, it was worth it. My advice to young traders out there is don't put a ceiling on what you can create. Try, fail, fail again, fail some more. Do it, until you succeed. The guys at Q are also helpful at times, and they're always glad to give guidance to newbies. It's been fun guys. It really has been. We should do a documentary when it's all said and done. A lot of bright people contributed here to make this happen. The folks at Q, John who created the platform, even the people who were critical. Everyone played a role. It's sort of ironic. This entire ordeal felt like this.

Hi @Aniken Leo,
Personally i hope you do make a lot of $$$ and retire in a year's time. It is a wonderful goal to have. I set a longer timeline than that for myself ... in my case it was 5 years rather than 1, ... AND then 1 had to repeat that goal again several times ... but anyway now at last i am retired, it's wonderful, and i wish you every success in your journey. Best wishes to you that you may get there faster than i did myself!

Now i have a question for you that relates to the fact that I am looking at Q's "new Risk Model", trying to make sense of it, and trying to figure out how to actually use it, but I am struggling with it. See the thread: "Risk Model Example: Detecting High Short Term Reversal Risk". In your post above, you quote Q's statement: "There are five investing styles we track: size, value, momentum, short-term reversal, and volatility." and then you go on to demonstrate your understanding of that. So my practical question to you, actually several closely related questions: Have you used Q's "new Risk model"? Do you understand it? Does it make sense to you? Can it actually be applied? If so, can you see how? If not, then what do you use as an alternative?

All I'm saying is, if you don't meet the requirements for the Quantopian fund, you shouldn't expect to do well in the competition.

To use an example:
You wouldn't apply to Cambridge University without having met their entry requirements, no matter how well you have done in a single exam. If you did and you didn't get in, that wouldn't make Cambridge a poor university, that would make you an arrogant applicant.

If you were to make your algorithm hedged and continue to do as well out of sample you may be in with a shot of winning. As it is however, you didn't meet the requirements and as a result, shouldn't expect to do well.

The requirements are quite clearly laid out here:
""""" Beta to SPY
The first change is that your algorithm is now scored according to how connected your algorithm's performance is to SPY. The lower your connection to SPY, the better. We already had 6 equal-weighted factors that generate your score; beta is now a 7th factor, all of them still equal-weighted.

Why did we do this change? If you look at a chart of SPY for February and March (the Quantopian Research notebook is attached), you could almost believe that the Quantopian Open was the driving factor in the S&P performance. On February 2, the start of the February judging period, SPY got on a rocket and headed for the stars. Grant's algo rode that rocket to the top of the charts, and he started trading real money on March 2. On March 2 the rocket ran out of fuel, and Grant's algo suffered! As we build the Quantopian hedge fund, we need to find algorithms that are uncorrelated from each other. The biggest correlation we're seeing today is around the S&P 500. This change is designed to encourage algorithms that are not correlated.

Consistency Between Paper Trading Results and Backtesting
The second change is that algorithms are now scored on how consistent they are between their paper trading returns and backtesting returns. The more consistent you are, the better. This factor is added to the calculation at the very end. After we compute what used to be referred to as your final score, we now multiply it by the consistency number, and the result is the new final score. This is applied gradually over the first few days of trading while the paper trading record is very volatile, and is fully applied at 20 days of trading.

We put in this change for a couple of reasons. The biggest reason is that we were seeing a lot of algorithms that had really good backtests that just weren't doing well in paper trading. This isn't too surprising when you think about it - if you're trying for a good score, you invest time in the backtest, and it's pretty easy to fall into data-mining, data-snooping, curve-fitting, or whatever you want to call that mistake. If you're prone to that mistake, you're not going to make an algorithm that lasts in the long run. We want to strongly encourage people to use good practices with out-of-sample data testing. If we make it very clear that a good backtest, on its own, can't win the contest, we hope to get more careful thought about how to write an algorithm that will perform well in paper trading.

The second reason is to discourage cheaters. We've seen a few instances where contest submissions are being deliberately gamed by submitting a "perfect backtest" and then a coin-flip over several entries for the paper trading. We've disqualified them, and we will continue to disqualify them in the future. The scoring change is a bit of a safety net, and a clear signal that it's not a strategy that will succeed.

For the detail-oriented: we're computing the consistency score using a kernel-density estimate using Gaussian kernels found in the Python scipy package. Both the backtest daily returns and the paper trade daily returns are each pushed through the function to fit them to a distribution separately. The difference between the areas of each of the distribution curves is used for the consistency score.

Future Changes
When we kicked off the Quantopian Open we promised that we would iterate and improve the contest. We don't think today's changes are the last word. There will be future scoring and rules changes as we think are necessary. We hope that you've found the previous discussions about scoring to be helpful; we certainly have. As always, we welcome and value your feedback about how we can make the contest better.

Although I wouldn't invest in an algorithm which only made $37 off of a $10,000,000 investment, I also wouldn't invest in an algorithm which has 28% volatility. It is up to Quantopian to protect their own backs, and I completely understand the choice to only accept algorithms within a very small band. If I were taking investors' money and putting it into an algorithm made by an untested developer, I would be putting it into the algorithm which has as low volatility as possible, whilst being hedged. I definitely wouldn't put it into an algorithm which has massive 28% fluctuations...

These are my entries, they meet the requirements and have been doing well so far. They both score higher than yours and do so by doing what Quantopian has asked for.

My advice to you is, if you want an allocation, add another line into your code along the lines of order_target_percent(sid(8554), -0.01) in order to meet the hedged requirement.

Aniken Skywalker has raised a lot of valid points in terms of basic investment principles and contest scoring flaws. I just want to highlight the major flaw in scoring the leader in contest 34/35 below:


The error here is in the calculation of Sharpe/Sortino Ratio which did not account for adjusting mean annual returns with risk free returns (1 yr. U.S. Treasury Bill = 1.62% last I checked). For reference Sharpe ratio is calculated this way:

Sharpe ratio = (Mean portfolio return − Risk-free rate)/Standard deviation of portfolio return

Had the Sharpe/Sortino Ratio been calculated correctly, the current leader would have a negative Sharpe/Sortino and would probably be ranked very low!
If you can't beat the risk free return, you're just making the broker rich! If you can't beat SP500 returns, then you're not creating alpha. High risks, high returns! Low risks, low returns! Maybe, somewhere in the middle is what Q should be looking for! It is really Q's prerogative to set the parameters of what they are looking for in a portfolio, PROVIDED THEY COMPUTE THE METRICS CORRECTLY!!!

Hey Tony,

I have not explored Q's new risk model, but at some point I will. I just took on a new role as a Sr. Developer, so I'm swamped at the moment with work. However, Quantopian is addictive, and this is just another new toy in a huge toolbox. There are a lot of things I would like to experiment with, ESPECIALLY, some of the paid packages. That said, I'm gonna have to lobby you guys once more on live trading. Why not sell it as a package? I'd be willing to pay as much as $100 a month, maybe even more, if you let me turn my stuff back on. I honestly don't like the way IB's API is setup, and I'm not a big fan on how QuantConnect is setup at the moment either. Tell John if he can do that for the Empire, the Republic will pardon his transgressions.

You were right Tony. I'm being watched and followed. What are the odds that four cars all left their trunks open. Clearly, this was to identify where I reside. Now, I have to shift my mindset to the dark side. Most people would be terrified. I'm actually offended that someone would be audacious enough to think that I was a sheep.

Well, @Aniken Leo, i can only say that i hope they do not vectorize you and beam you off into the deeper parts of cyberspace, to be lost to the Q community forever. Everyone would deeply miss your commentary. So relax & stay safe!

Well, that would suck, especially if the algorithm has a sudden huge DD and they don't have the capacity to fix it. Also, in a life and death situation, I would probably just give it to them. Duh! Nevertheless, I think I'm worth more alive than dead, but hey, we all gotta go sometime. Today is as good a day as any. In the meanwhile, I had an idea that came to me yesterday with regards to options that I'm really excited to try. However, I need to do a couple days of research. I may be able to achieve even greater gains than this, with tremendously less risk. Then I'll need a clone army for real.