it is related to ai because otherwise profit factor per trade is not available on any trading platform - you can get profit factor for all trades in a series of trades a system makes - this is after the fact but to get real time profit factor is valuable tracking information to know as it's happening how you system is performing. sorry you and no one else this forum see the value
i have found that profit factor is the best fitness test for a model, but unfortunately its hard to get on a per trade basis. the average is basically "average(profit factor, trade count)" so as the trade count increases so does the average length just keeps going and becomes a stable average as you see in the graph. the code produces a file that looks like this Date,Time,ProfitFactor,RollingPF,TradingDays,TradeCount,MinTarget 2025-04-30,08:57:00,1.75,1.75,1,1,1.75 2025-05-01,10:46:00,2.50,2.13,2,2,1.75 2025-05-05,10:35:00,1.39,1.88,3,3,1.75 2025-05-06,11:14:00,2.78,2.10,4,4,1.75 2025-05-08,09:19:00,1.79,2.04,5,5,1.75 2025-05-09,10:08:00,1.37,1.93,6,6,1.75 2025-05-12,08:54:00,1.14,1.82,7,7,1.75 2025-05-13,10:27:00,1.73,1.81,8,8,1.75 2025-05-15,11:10:00,2.31,1.86,9,9,1.75 2025-05-19,09:07:00,2.89,1.96,10,10,1.75 2025-05-21,09:34:00,2.45,2.01,11,11,1.75 2025-05-22,09:23:00,1.99,2.01,12,12,1.75 2025-05-23,08:47:00,1.79,1.99,13,13,1.75 2025-05-27,08:42:00,1.62,1.96,14,14,1.75 this is what the file needs to look like for the html to read and update btw sincerely thank you for your interest
Exactly this. This was on my to do list!!!!! Thanks again. What is nice about this contribution it's completely separate from any strategy, but still interesting!
i am creating an expansion of this that will compare the profit factor of up to 10 models and then it will create a heatmap so you know which system is hot and which is cold. then i will add in contract allocation into the mix. without ai it would be a daunting task to code and debug all this is my point.
I'm not so sure it is desperation. It is a logical step. Once there is a measurement for "how it is doing", one can drill down and see if the information is orthogonal enough to merit it as a guide for model selection. Of course it will have the standard issues: Context, time frame, etc. But that should not prevent one from having the measurements even if they are not actionable right away. Often if the measurement is valid adding another valid measurement or two can make it sound. OTH it would be a reasonable metric to compare to other "like-kind" system, to see if it indeed different or worth the effort based on the amount of difference. This is standard process improvement stuff. I am out of this thread, as my sense it will decay from here....