Neural Network strategy backtest comparison
#1
Lightbulb 
Hi, all!

For the past few months, I've been running backtest experiments where I compare publicly available strategies against each other. I have been documenting results in my Medium series - https://medium.com/@deandree. I just thought I share this here, because I believe my results and information could be useful to Gekko community. Also I hope to hear insights and comments from you.

In my latest article, I decided to go for the big fish and gathered 15 Neural Network strategies and backtested them against TOP20 Coins on September datarange:
    -    neuralnet_SirTificate
    -    neuralnet_zschro
    -    zuki_nn
    -    mounirs-ga-version-1
    -    mounirs-ga-version-2
    -    mounirs_esto
    -    n8
    -    n8_v2
    -    n8_v2_BB_RSI_SL
    -    NNv2
    -    NN_ADX_RSI
    -    neataptic
    -    LSTM_MACD_RSI_V3
    -    Luke_NN
    -    ManuNet

In my next article, I have planned to dig deeper and show you the price prediction line of each strategy, so we can get a better understanding of what's really going on behind the scenes. 

For example, here is prediction line for n8_v2 strategy for IOT coin in September month:
[Image: n8-v2-IOT.png]

Let me know if this is useful to you and/or if any suggestions! Thanks!
  Reply
#2
Thumbs Up 
thank you   

Very nice and detailed review thanks

https://medium.com/@deandree/can-neural-...547b5fe415


do you consider the time factor

the minute you start the strategy is very important


(00) 
maybe lost in minutes

(01) maybe you can profit in minutes
  Reply
#3
Yes, starting time is very important. I've seen quite huge (x2/x3) difference on 10m candles, on 120/240/etc it could be much higher.

Due to already long time it takes to run my tests for each Medium part, I have not included multiple starting time in my tests. If I wanted to do that, I could try a few offsets (like for 240m candles take 0h 1h 2h 3h offset) and calculate average. If people want to see that, I can make special article about that.
  Reply
#4
Hey, all!

For those of you who enjoyed Part 4 about NNs, I have released 2 new parts in last week:

Part 5 - extension to part 4, where I backtest NN strats against October, August and July. I also compare how to market itself was doing, so we can keep the results in context. For example, here are results for July 240m:

[Image: 240m-Candles.png]



Part 6 - I get started on prediction analysis by trying to predict single event - huge price drop that started from November 14th. I draw the prediction lines on chart, so we can compare the accuracy, for example, here is neataptic strategy predictions with 5m candles and 900 history:

[Image: neataptic.png]

Let me know if this works for you and if any suggestions for future topics. Thanks!
  Reply
#5
This is very cool stuff!

Quote:For example, here is prediction line for n8_v2 strategy for IOT coin in September month:

For me that image in your OP isn't loading.

I will be watching this thread and can hopefully contribute some day as well Smile
  Reply
#6
@askmike thanks, I fixed the image in first post.
  Reply


Forum Jump:


Users browsing this thread: