Book: http://www.amazon.com/Stocks-Move-Beating-Momentum-Strategies/dp/1511466146

Website: http://www.followingthetrend.com/

This is my attempt at a faithful recreation of his system. Sometimes it dips into margin a little, I haven't isolated why. Comments welcome!

Also, due credit to Ted, who shared another implementation here: https://www.quantopian.com/posts/anyone-found-a-substantial-momentum-effect (and James Christopher who says he's done one too).

Clone Algorithm

1637

Loading...

There was an error loading this backtest.

Backtest from
to
with
initial capital

Cumulative performance:

Algorithm
Benchmark

Custom data:

Total Returns

--

Alpha

--

Beta

--

Sharpe

--

Sortino

--

Max Drawdown

--

Benchmark Returns

--

Volatility

--

Returns | 1 Month | 3 Month | 6 Month | 12 Month |

Alpha | 1 Month | 3 Month | 6 Month | 12 Month |

Beta | 1 Month | 3 Month | 6 Month | 12 Month |

Sharpe | 1 Month | 3 Month | 6 Month | 12 Month |

Sortino | 1 Month | 3 Month | 6 Month | 12 Month |

Volatility | 1 Month | 3 Month | 6 Month | 12 Month |

Max Drawdown | 1 Month | 3 Month | 6 Month | 12 Month |

# Implementation of the Stocks On The Move system by Andreas Clenow # http://www.followingthetrend.com/stocks-on-the-move/ import numpy as np import pandas as pd import scipy.stats as stats import time from quantopian.pipeline import Pipeline from quantopian.pipeline import CustomFactor from quantopian.pipeline.data import morningstar from quantopian.pipeline.factors import EWMA, Latest, SimpleMovingAverage from quantopian.pipeline.data.builtin import USEquityPricing from quantopian.algorithm import attach_pipeline, pipeline_output UniverseSize = 500 DailyRangePerStock = 0.001 # targeting 10bp of account value RebalanceThreshold = 0.005 # don't rebalance if the difference is less than 50bp of account value # This is the momentum factor, which is the 'slope' of an exponential regression # (ie a linear regression of logarithms), multiplied the the R-Squared of that regression class Momentum(CustomFactor): inputs = [USEquityPricing.close] window_length = 90 def compute(self, today, assets, out, close): x = pd.Series(range(0,self.window_length)) log_close = np.log(close) scores = np.empty(len(close.T), dtype=np.float64) for i in range(0,len(assets)): if (not np.all(np.isnan(log_close[:,i]))): y = np.copy(log_close[:,i]) # interpolate NaN, not forward-looking since we are regressing anyway try: mask = np.isnan(y) y[mask] = np.interp(np.flatnonzero(mask), np.flatnonzero(~mask), y[~mask]) slope, _, r, _, _ = stats.linregress(x, y) scores[i] = slope * 256.0 * r * r except: scores[i] = -1000.0 log.error("Regression error!") else: scores[i] = -1000.0 out[:] = scores def descending_rank(a): return a.argsort()[::-1].argsort() # This is the momentum factor, except only calculated for those stocks which will end up # keeping within our set_screen. class MomentumOfTopN(CustomFactor): inputs = [USEquityPricing.close, morningstar.valuation.shares_outstanding] window_length = 90 def compute(self, today, assets, out, close, shares): # get our universe in here again because lame starting_caps = close[-1] * shares[-1] starting_caps[np.isnan(starting_caps)] = 0.0 cap_ranks = descending_rank(starting_caps) x = pd.Series(range(0,self.window_length)) log_close = np.log(close) scores = np.empty(len(close.T), dtype=np.float64) for i in range(0,len(assets)): if (cap_ranks[i] < UniverseSize): if (not np.all(np.isnan(log_close[:,i]))): y = np.copy(log_close[:,i]) # interpolate NaN, not forward-looking since we are regressing anyway try: mask = np.isnan(y) y[mask] = np.interp(np.flatnonzero(mask), np.flatnonzero(~mask), y[~mask]) slope, _, r, _, _ = stats.linregress(x, y) scores[i] = slope * 256.0 * r * r except: scores[i] = -1000.0 log.error("Regression error!") else: scores[i] = -1000.0 else: scores[i] = -1000.0 out[:] = scores class MarketCap(CustomFactor): inputs = [USEquityPricing.close, morningstar.valuation.shares_outstanding] window_length = 1 def compute(self, today, assets, out, close, shares): out[:] = close[-1] * shares[-1] # the biggest absolute overnight gap in the previous 90 sessions class MaxGap(CustomFactor): inputs = [USEquityPricing.close] window_length = 90 def compute(self, today, assets, out, close): abs_log_rets = np.abs(np.diff(np.log(close),axis=0)) max_gap = np.max(abs_log_rets, axis=0) out[:] = max_gap # the Average True Range over the last 20 sessions class ATR(CustomFactor): inputs = [USEquityPricing.close,USEquityPricing.high,USEquityPricing.low] window_length = 21 def compute(self, today, assets, out, close, high, low): hml = high - low hmpc = np.abs(high - np.roll(close, 1, axis=0)) lmpc = np.abs(low - np.roll(close, 1, axis=0)) tr = np.maximum(hml, np.maximum(hmpc, lmpc)) atr = np.mean(tr[1:], axis=0) out[:] = atr def initialize(context): context.spy = sid(8554) set_benchmark(context.spy) momentum = MomentumOfTopN() mkt_cap = MarketCap() max_gap = MaxGap() atr = ATR() latest = Latest(inputs=[USEquityPricing.close]) mkt_cap_rank = mkt_cap.rank(ascending=False) universe = (mkt_cap_rank <= UniverseSize) momentum_rank = momentum.rank(mask=universe, ascending=False) sma100 = SimpleMovingAverage(inputs=[USEquityPricing.close], window_length=100) pipe = Pipeline() pipe.add(momentum, 'momentum') pipe.add(max_gap, 'max_gap') pipe.add(mkt_cap, 'mkt_cap') pipe.add(mkt_cap_rank, 'mkt_cap_rank') pipe.add(sma100, 'sma100') pipe.add(latest, 'latest') pipe.add(atr, 'atr') pipe.add(momentum_rank, 'momentum_rank') # pre-screen all the NaN stuff, and crop down to our pseudo-S&P 500 universe pipe.set_screen(universe & (momentum.eq(momentum)) & # these are just to drop NaN (sma100.eq(sma100)) & (mkt_cap.eq(mkt_cap)) ) pipe = attach_pipeline(pipe, name='sotm') # do our work on Wednesdays, as in the books. schedule_function(func=allocate_1, date_rule=date_rules.week_start(days_offset=2), time_rule=time_rules.market_open(minutes=60), half_days=True) schedule_function(func=allocate_2, date_rule=date_rules.week_start(days_offset=2), time_rule=time_rules.market_open(minutes=90), half_days=True) schedule_function(func=allocate_3, date_rule=date_rules.week_start(days_offset=2), time_rule=time_rules.market_open(minutes=120), half_days=True) schedule_function(func=record_vars, date_rule=date_rules.every_day(), time_rule=time_rules.market_open(minutes=1), half_days=True) schedule_function(func=record_vars, date_rule=date_rules.every_day(), time_rule=time_rules.market_close(), half_days=True) schedule_function(func=cancel_all, date_rule=date_rules.every_day(), time_rule=time_rules.market_close(), half_days=True) context.rebalance_needed = False set_slippage(slippage.FixedSlippage(spread=0.01)) set_commission(commission.PerShare(cost=0.0035, min_trade_cost=0.35)) # This function trims down the Pipeline results to those stocks which allow our portfolio # to hold. Stocks which fall out of these criteria are sold. # 1. In the top 20% of the stocks, ranked by their momentum # 2. No recent gaps more than 15%, in either direction # 3. Stock is trading above it's 100-session moving average. Here we use the exponential # moving average, I think the book might be using the simple moving average. def filter_pipeline_results(results): # remove those whose momentum rank is not in the top 20% filtered = results[results['momentum_rank'] < 0.2*UniverseSize] # filter out gaps filtered = filtered[filtered['max_gap'] < 0.15] # filter out stocks under 100 EMA filtered = filtered[filtered['latest'] > filtered['sma100']] return filtered def before_trading_start(context, data): results = pipeline_output('sotm').sort('momentum_rank') filtered = filter_pipeline_results(results) context.pool = filtered update_universe(filtered.index) def sell_positions(context, data): cash_freed = 0.0 s = "" for sid in context.portfolio.positions: position = context.portfolio.positions[sid] cash_worth = position.amount * position.last_sale_price # anything not in the pool of allowed stocks is immediately sold if ((sid not in context.pool.index) & (sid in data)): s = s + "%s, " % sid.symbol order_target_percent(sid, 0.0) cash_freed = cash_freed + cash_worth log.info(s) return cash_freed def desired_position_size_in_shares(context, data, sid): account_value = context.account.equity_with_loan target_range = DailyRangePerStock estimated_atr = context.pool['atr'][sid] return (account_value * target_range) / estimated_atr def rebalance_positions(context, data): account_value = context.account.equity_with_loan cash_freed = 0.0 s = "" for sid in context.portfolio.positions: position = context.portfolio.positions[sid] current_shares = position.amount if (sid in context.pool.index): target_shares = desired_position_size_in_shares(context, data, sid) sid_cash_freed = (current_shares - target_shares) * position.last_sale_price # only rebalance if we are buying or selling more than a certain pct of # account value, to save on transaction costs if ((abs(sid_cash_freed / account_value) > RebalanceThreshold) & (sid in data)): s = s + "%s (%d -> %d), " % (sid.symbol, int(current_shares), int(target_shares)) order_target(sid, target_shares) cash_freed = cash_freed + sid_cash_freed log.info(s) return cash_freed def should_rebalance(context): ret = context.rebalance_needed context.rebalance_needed = not context.rebalance_needed return ret # This returns the global switch as to whether we can add any new positions, # or only sell/rebalance positions. def can_buy(context, data): latest = data[context.spy].close_price h = history(200,'1d','close_price') avg = h[context.spy].mean() return latest > avg # This function is for adding new positions, by iterating through the # eligible stocks in order of momentum, and buying them if we have (anticipate # having) enough cash to do so. def add_positions(context, data, cash_available): s = "" for i in range(0,len(context.pool)): sid = context.pool.index[i] if ((sid not in context.portfolio.positions) & (sid in data)): desired_shares = desired_position_size_in_shares(context, data, sid) cash_req = desired_shares * data[sid].close_price if ((cash_req < cash_available)): s = s + "%s (%d shares), " % (sid.symbol, int(desired_shares)) order_target(sid, desired_shares) cash_available = cash_available - cash_req log.info(s) def allocate_1(context, data): log.info("Selling...") cash_from_sales = sell_positions(context, data) def allocate_2(context, data): if (should_rebalance(context)): log.info("Rebalancing...") cash_from_rebalance = rebalance_positions(context, data) def allocate_3(context, data): if (can_buy(context, data)): log.info("Buying...") add_positions(context, data, context.portfolio.cash) else: log.info("Cannot buy, pass.") def record_vars(context, data): record(PctCash=(context.portfolio.cash / context.account.equity_with_loan)) record(CanBuy=can_buy(context, data)) pos_count = len([s for s in context.portfolio.positions if context.portfolio.positions[s].amount != 0]) record(Stocks=(pos_count / 100.0)) # scale so that the other numbers don't get squished def handle_data(context, data): pass def cancel_all(context, data): sids_cancelled = set() open_orders = get_open_orders() for security, orders in open_orders.iteritems(): for oo in orders: sids_cancelled.add(oo.sid) cancel_order(oo) n_cancelled = len(sids_cancelled) if (n_cancelled > 0): log.info("Cancelled %d orders" % n_cancelled) return sids_cancelled