Excellent question. If the strategies are cross sectional and trade many assets, then things get easier. This means that they're usually going to rebalance once per time period (day, week, etc.), and all the individual strategies or models can be combined at a weighting level before portfolio optimization.
When combining cross sectional signals or models which produce weights, the general question you ask is the right one. Is the new model actually adding anything new? Is it a linear combination of my existing ones? You can think about it this way. Every model you're already trading is now beta, and you want to be producing alpha in your new model. You can check this by regressing your existing model returns (independent variables) against your new model (dependent variable), and seeing what percentage of the variance is explained.
If there's a significant amount of alpha, then you likely want to include your new model and average it in in some way.
If there's no alpha, look if the new model is mostly correlated with one or many existing models. If it's one you may want to swap the two or average them together into one new composite model. If it's correlated with many, then it's trickier.
In general more models are better, so I suspect in many circumstances you're gonna want to include your new model in to help diversify your model pool. The key is just to adjust weighting accordingly so as not to over-lever certain models. This can be accomplished by reducing the weights on the correlated models.
If the models don't trade regularly then it gets a little trickier. In that case you can't really combine at a weighting level. You need to understand how all the strategies are correlated by doing a regression as mentioned earlier, and then apply a top level constrained portfolio optimization. This means that if a strategy tries to purchase a stock that would violate a position concentration constraint, the optimizer won't let it. Obviously sometimes you can catch this by looking at historical performance and deciding that two models will just be doing the same thing, but if they're uncorrelated they may still sometimes take over-risked positions. Having a top level constrained optimization before any trade is still very important.
Here are some additional resources which may be helpful when thinking about this:
The material on this website is provided for informational purposes only and does not constitute an offer to sell, a solicitation to buy, or a recommendation or endorsement for any security or strategy, nor does it constitute an offer to provide investment advisory services by Quantopian. In addition, the material offers no opinion with respect to the suitability of any security or specific investment. No information contained herein should be regarded as a suggestion to engage in or refrain from any investment-related course of action as none of Quantopian nor any of its affiliates is undertaking to provide investment advice, act as an adviser to any plan or entity subject to the Employee Retirement Income Security Act of 1974, as amended, individual retirement account or individual retirement annuity, or give advice in a fiduciary capacity with respect to the materials presented herein. If you are an individual retirement or other investor, contact your financial advisor or other fiduciary unrelated to Quantopian about whether any given investment idea, strategy, product or service described herein may be appropriate for your circumstances. All investments involve risk, including loss of principal. Quantopian makes no guarantees as to the accuracy or completeness of the views expressed in the website. The views are subject to change, and may have become unreliable for various reasons, including changes in market conditions or economic circumstances.