Welcome to NexusFi: the best trading community on the planet, with over 150,000 members Sign Up Now for Free
Genuine reviews from real traders, not fake reviews from stealth vendors
Quality education from leading professional traders
We are a friendly, helpful, and positive community
We do not tolerate rude behavior, trolling, or vendors advertising in posts
We are here to help, just let us know what you need
You'll need to register in order to view the content of the threads and start contributing to our community. It's free for basic access, or support us by becoming an Elite Member -- see if you qualify for a discount below.
-- Big Mike, Site Administrator
(If you already have an account, login at the top of the page)
I have spent last three weeks developing an auto strategy since I am not fast and decisive enough to trade discretionary these times. NQ is three times faster than it used to be before the holiday...
Anyway, I tried backtesting the strategy on 2 months of historic data and fixed the strategy enough that I arrive at similar results in backtest and replay.
Despite I managed to decrease the drawdown and improved the performance for June, I am still in doubt whether the strategy is fit enough to run be run live. I understand that the strategy needs to be optimized from time to time, but the market was very different in June than in July and I don't know if it is necessary to fit to very different market conditions.
The attached images are the results of the backtest. The numbers are for one contract trading (I don't have account big enough to try to scale yet).
Any thoughts? Would any of you consider this good enough? The expectancy BTW is far better than in my discretionary trading.
Thanks
Petr
Can you help answer these questions from other members on NexusFi?
On your chart, the first half of trading is a net loss. Then, suddenly it became hugely profitable in the second half - are you sure this isn't just curve fit to the second half of the trading period? Why not backtest it over a year or 2 years of data?
@pengi, it is probably curve fitted. I run it on the couple of years and the results are not as good. But the thing is that the market conditions have changed quite recently. How to take that into account? I would need to find the similar conditions in the years back...
@kojava, I think that the slippage of two ticks is a bit too harsh. The strategy is able to withstand one tick slippage, but two ticks are killer.
Anyway, I gave it a try live with one contract.
today +$135, yesterday +$120, Friday -$140... everything in total.
not too good.
In the backtest, the strategy has positive expectancy for the whole year of data, but it nets far too little.
I am not able to eliminate the trades in the times the market goes sideways or ranging.
So I could only use it in semi-automatic mode when I could turn it off, if I don't like the conditions.
In my experience, I've found that my strategy should have a positive expectancy and a net gain every day before ever running the optimizer. If it doesn't, then my entry and exit code probably isn't solid. The optimizer should only tweak your numbers slightly, to avoid curve fitting your strategy (ie. destroying it).
I try to avoid using variables that are directly related to market conditions that change over time. My good strategy should be profitable all year, every month, every week, every day, and every hour. If it isn't, then it's not a good strategy!
I have recently found another strategy that I could automate.
Here are the results on 4 months of data. It performs better than my previous strategies but still not sure, whether it is fit enough.
The drawdown is big and it is certainly not profitable every day. The winners are on average 2x bigger than losers.
Commissions calculated, slippage standard.
I tried to simulate with different starting capital (10k) and the results were similar (25k), but the return was not so big in percentage.
The exits are something I need to work on - I am leaving money on the table, so I need to perhaps implement some better exit strategies.
Is there some way to improve the equity curve other than strategy portfolio diversification?
Although I am not really in the automation corner, I at least have been there ( my quest for the holy grail, but the closest I could find was price action).
So for my taste the drawdowns are very large, run a MC over that data to see what kind of drawdowns you get there.
Also I would suspect due to the shape of the p/l curve that your bot doesn't cope too well with changing market conditions, but that's just a guess.
An other hint toward that would be if the MC wold suggest much smaller drawdowns as the ones observed.
thanks for hints. Maybe stupid question, but what the MC acronym stands for?
The second point is very true. NQ has changed a lot (much more volatile) and the bot works better in bigger volatility. I need to decrease the number of trades in worse conditions and also decrease the Profit targets.
I have just tried that and the bot is mostly break-even in such situation.
Once the volatility pops, the fixed ratio kicks in and the curve flies high....