I dont know whats going on. I have an algo I paid for from someone. It’ll show 75% profitable trades in the last 90 days but then something changes and it is no longer profitable in the same time period. Its profitable for their group, I’m the only one with the problem. Then it even fails in live testing. It happens on all strategies. I redownload historical tick data to make sure its not corrupted or something.
You got played. Maybe the group is in on it? If you have an algo and using it as them then you should have similar results in theory. It’s very easy to make profitable looking algos based on backtest. I’ve made 99.99% winning algos based on a couple years of backtest before. It lost maybe two times out of thousands of trades. Stop paying for those scams. Nobody is going to be selling real profitable algos.
3 Likes
I didnt say years, I said 90 days, which is reliable. I too can make strategies that have 30 profit factor or whatever. What you said is not the case. It happens on my strategies too. I’m saying its showing me different results. Right now its looking correct but usually it changes to very bad results. It shouldnt change like that from one moment to the next.
If it happens on your strategy too then just accept that both algos are trash. You paid for a trash strategy.
Have you identified what has changed?
I have seen this behavior as well. And unfortunately I don’t have a solution/answer for you. I have a strategy I wrote myself which logs all my trades to a database. I have run this strategy on live data on my live account, then at the end of that day, run a backtest (Analyzer) with the same data. Then download the replay file and run again at normal speed. 3 wildly different results, presumably using the exact same data. I understand that I may not be getting the EXACT same ticks in a bar of data from real time to historical, but should be reasonably close. My deep dive showed that the events aren’t always firing in the same sequence when backtesting (OnBarUpdate, OnExecutionUpdate, OnOrderUpdate). Nothing I can do about that, other than try to handle the obvious exceptions in code. Sorry, wish I had an easy answer for you. Just know that results of a backtest are only an approximation. Gotta run your algo real time against a SIM account for several weeks before lighting the fuse with real money.
1 Like
You’ll see that behavior usually with tick derived bars like tick and volume bars. Time series and range seems ok. The issue is that the tick may slightly change every time you run a backtest or replay, which you can verify by just looking at the chart. If you replay something like a tick or volume chart, you’ll notice that they are not the same all the time. If you have derived metrics, this compounds and you eventually have significantly different values.
1 Like
3 minute chart just clicking on historical history