Thread by @mikeharrisNY

Thread about trading strategies. Get a bowl of rice and chopsticks but never stick the chopsticks into the rice and never cross them. Those are bad signs.

I started developing #tradingstrategy in the early 90s accidentally.

I was working in Wall Street in fixed income developing algos for bond portfolio management, calculating funding gaps for money market desks and trading the bond basis. They were interesting but boring jobs.

Someone I knew asked me if I could backtest a strategy for trading currency futures. I accepted the challenge but it turned out the strat wasn't any good. But I found the trick to make it good. The person insisted this was a way to "print" our own money. I had my doubts.

We invested a large sum of our own money and traded the strat for less than a year making a significant profit. My partner by then insisted we start a fund. I never wanted to trade OPM. I told him we were probably lucky, took my cut and run as fast as I could.

I realized having a strat isn't nearly enough. It must be robust and not "lucky" outcome. Then that strat was trend-following. The discipline required was horrendous. I was in charge of trading and I couldn't sleep at night. There was no diversification. Something could go wrong.

I saw potential in trading strategies but I decided looking for short-term price action anomalies and "hit and run" trading was more suitable for me. Then I started on the long path of looking for an edge using everything I new from math, probability and SITG.

End of introduction. Trading strategies pose significant tradeoffs. Two of them are: CAGR vs. max DD and Profit Factor vs. win rate. Finding the proper balance for these parameters is a daunting task. As you try to increase win rate and CAGR, the PF and DD go down, normally.

Some people say: "You don't need high win rate as long as PF is high" Those are maybe fools. Win rate is tied to probability of ruin, the lower the win rate the higher the probability of a long streak of losers. You need sufficiently high win rate to survive.

Then as you try to get more juice from the market you open up yourself to higher drawdown. I see people discussing strategies with 40% or even 50% maximum drawdown historically. No trader with SITG will consider anything above 20%. Funds often close down after 20% to 25% drawdown

More importantly, I see people discussing strategies with < 0.6 Sharpe. They better become passive investors. $SPX buy and hold Sharpe is about 0.6. If your Sharpe is less than at least 0.75, hitting uncle point is guaranteed at some point. Ignore the low Sharpe strats.

Then, having one strat with Sharpe of 1 is not nearly enough. You need diversification. It is better to have two uncorrelated strategies with 0.7 Sharpe and they will give you a portfolio Sharpe close to 1 but also protection from correlated markets. The game is hard.

Ignore talk about non-stationarity and non-ergodicity. These are important but often philosophical debates with no value to SITG #traders. Academics may get tenure credits discussion those but traders get P/L credits only. Non-stationarity is what makes you money.

In fact, trading strategies attempt to extract profits from non-stationary markets. They do that with timing and risk management. Those who claim that non-stationarity is problem for trading strategies are largely confused. Non-stationarity IS the source of profits.

Non-ergodicity is mostly a philosophical debate. Ignore any attempts to shift the attention of research to this area as you will get nowhere. It's an interesting subject but will serve as a distraction from the task of developing an edge.

Developing trading strategies in a long, steep learning curve. Knowing when to abandon a strategy is also part of the edge. This subject is rarely discussed because is close to alchemy. Simple rule I use: if max DD makes new high, I abandon.

Beware of strategies suggested by data. They may be good but usually are so over-fitted than no validation test can prove that. On the other hand, strategies based on unique hypotheses are difficult to conceive. Those who manage that get a license to "print money".

Strategies suggested by the data (via ML, GP, NN, etc.) are usually Type-I errors (false positives) Never pay attention to academics that claim there are tests for identifying the error. There is none and all those are attempts to get some tenure credit IMO.

Unique hypotheses may not be unique since there are thousands looking for them every day. If you think you have a unique hypothesis, the first task should be to try to debunk it instead of getting too excited about it.

Finally, I have said in the past, trading became difficult during the 90s and there was a significant regime change in markets primarily because too many astute mathematicians and physicists joined the game after the fall of communism. Then the Chinese joined. The game is hard.

Conclusion: You need strategies with high Sharpe, diversification and as high win rate as possible to reduce probability of long losing streaks. If these conditions are not present, hitting uncle point is nearly guaranteed.