search

freqtrade / freqtrade

Free, open source crypto trading bot
https://www.freqtrade.io
GNU General Public License v3.0
27.99k stars 6k forks source link

have not enough space for profit field! #4654

Closed mablue closed 3 years ago

mablue commented 3 years ago

My environment:

The problem:

1) have not enough space for profit field in hyperopt results 2) Is this real? If yes, I will be a reach man after 1 mounth! :D

Observed Results:

+--------+-----------+----------+------------------+--------------+-------------------------------+----------------+-------------+                                                                                         
|   Best |     Epoch |   Trades |    Win Draw Loss |   Avg profit |                        Profit |   Avg duration |   Objective |
|--------+-----------+----------+------------------+--------------+-------------------------------+----------------+-------------|
| * Best |    2/1000 |       12 |      3    4    5 |       -1.07% |  -47.44376800 USDT  (-12.81%) |        650.0 m |     2.73358 |
| * Best |    3/1000 |       45 |     18    1   26 |        0.85% |  141.48183841 USDT   (38.20%) |        329.3 m |    -3.21775 |
| * Best |    8/1000 |      279 |    151   81   47 |        4.91% | 5,078.13619263 USDT (1,371.10%) |      2,290.5 m |    -57.4013 |
| * Best |    9/1000 |      723 |    333   17  373 |        1.33% | 3,556.15266428 USDT  (960.16%) |        815.7 m |    -76.0623 |
| * Best |   16/1000 |     2410 |   1665   66  679 |        1.34% | 11,954.11413026 USDT (3,227.61%) |        383.9 m |    -338.614 |                                                                                      
|   Best |   54/1000 |     1276 |   1028  187   61 |        2.60% | 12,288.71973612 USDT (3,317.96%) |        906.3 m |    -348.458 |                                                                                      
|   Best |  158/1000 |     1277 |   1027  151   99 |        3.58% | 16,926.48323613 USDT (4,570.16%) |      1,018.6 m |    -373.872 |                                                                                      
|   Best |  186/1000 |     4538 |   3371   25 1142 |        1.19% | 19,970.90026418 USDT (5,392.15%) |        287.3 m |    -658.312 |                                                                                      
|   Best |  222/1000 |     4872 |   3558  166 1148 |        1.23% | 22,172.65932511 USDT (5,986.62%) |        270.8 m |    -769.218 |                                                                                      
|   Best |  280/1000 |     4767 |   3384   31 1352 |        1.45% | 25,596.11192869 USDT (6,910.96%) |        297.2 m |    -829.461 |                                                                                      
|   Best |  294/1000 |     8495 |   6747   39 1709 |        1.44% | 45,463.75116653 USDT (12,275.23%) |        224.9 m | -1,563.04204 |                                                                                    
|   Best |  741/1000 |     8235 |   7113   44 1078 |        1.68% | 51,118.77908033 USDT (13,802.08%) |        193.8 m | -1,607.13968 |                                                                                    
|   Best |  779/1000 |    13398 |  10101  200 3097 |        1.00% | 49,414.73272481 USDT (13,341.99%) |        120.3 m | -1,805.90920 |
hippocritical commented 3 years ago

Hi, what strategy are you using ?

Veeeery likely this won't happen in reality since the Backtesting does not wait for offers to get filled by the server and other things. I would recommend running some days in dry mode then you will see pretty fast if the result is even remotely realistic.

The new strategy of mine has a e^19 profit within 3 months ... That can't be realistic either. Amazing would be 2% per day but on average even that is very hard to reach.

If your strategy is using future data then no wonder why it works like a charm. Look on freqtrade.io to strategy pitfalls of you want to investigate.

xmatthias commented 3 years ago

Please read the backtesting assumptions. Most likely, your strategy is exploiting one of the assumptions - and if you enable compounding, the "later" trades will have unrealistic trade sizes (several million per trade) which will NEVER fill on an exchange.

Also - just posting the hyperopt result will not help in even trying to investigate - for this we'll need the hyperopt file + the detailed results.

@mablue As you're spamming the issues throughout the freqtrade org, please let me be clear Please only post an issue if you actually encounter a problem, or really have a question that is not answered in the documentation.

Looking and answering at pointless issues (like this one) is wasting people's time (it's people's free time, as this is a free, open source project), which would be better spent elsewhere, thanks.

If you need someone to talk, bests join the freqtrade discord server - where you can discuss your results with the community.

mablue commented 3 years ago

Yes yes sorry I know maybe this issues not helpful for this project.I just want to show the problem of the hyperopt result that you can see objective is not good indicator for finding best results and i tested in dry it cant get profit and no it have not generate big thrade sizes no combondig enabled.I just write a hypropt an run it and see this amazing result.sorry for my badless

mablue commented 3 years ago

Hi, what strategy are you using ?

Veeeery likely this won't happen in reality since the Backtesting does not wait for offers to get filled by the server and other things. I would recommend running some days in dry mode then you will see pretty fast if the result is even remotely realistic.

The new strategy of mine has a e^19 profit within 3 months ... That can't be realistic either. Amazing would be 2% per day but on average even that is very hard to reach.

If your strategy is using future data then no wonder why it works like a charm. Look on freqtrade.io to strategy pitfalls of you want to investigate.

Yes it not happend in real.so what can we do?we need an indicator to find best iteration of the hyperopt and developer closing the issues! :|

mablue commented 3 years ago

I fixed the problem : My config file configed to make unlimit trades with increasing the mony and stake size is a static number that this make huge trades in on baktest and hyperopt. To fix the problem i changed max trade from -1 to 10 and changed stake size to unlimited

xmatthias commented 3 years ago

we need an indicator to find best iteration of the hyperopt and developer closing the issues! :|

Freqtrade is not a "turn on and get rich" program. You will need to spend time and effort both developing a strategy and then testing and finetuning it to make it suit your preferences and risk profile.

If you're looking for indicators, please use google / any other search engine and research the different indicators - what their different signals mean and how to use them in trading. we will not provide support on these things - and expect a certain level of self-sustainment and self-research.

mablue commented 3 years ago

we need an indicator to find best iteration of the hyperopt and developer closing the issues! :|

Freqtrade is not a "turn on and get rich" program. You will need to spend time and effort both developing a strategy and then testing and finetuning it to make it suit your preferences and risk profile.

If you're looking for indicators, please use google / any other search engine and research the different indicators - what their different signals mean and how to use them in trading. we will not provide support on these things - and expect a certain level of self-sustainment and self-research. No I'm not mean technical indicators I mean dicreasing of objective field of hyperopt result.that some times(when trades count growing up)not show real best result. But yes you can not do anything about that filling/nit filling of trades in backtest or hyperopt is not have a formula in real exchenges. I dont mean you shoud fix it.It will happend cuz of: (max trade number : -1)in config file...