8 Ways to Improve your Backtesting and Optimization Process | Trading Strategy Development

The processes of backtesting and optimizing are difficult to get right, with one of the main issues being over-fitting. This video highlights 8 areas that can be used to improve the effectiveness of the process including:
Identifying a true trading edge
- Improving statistical significance and sample size
- The dangers of over-fitting
- How to avoid over-fitting
- Coding a multi-symbol expert advisor in MQL5
- Optimization design and walk forward phases
- Optimization profiles best-practice
- Performance metrics

Follow Darwinex on LinkedIn:

Further videos covering more detail of the issues raised in this video can be found here:

#BacktestingATradingStrategy, #TradingOptimizations, #OverFitting, #Backtesting, #TradingEdge, #StatisticalSignificance, #SampleSize, #NumberOfTrades, #HowToAvoidOverFitting, #MultiAssetTradingAlgo, #InSample, #OutOfSample, #WalkForward, #PerformanceMetrics, #ReturnDrawdown, #ProfitFactor, #SharpeRatio, #CalmarRatio, #AlgoTrading, #MartynTinsley, #Darwinex

Video Contents:
00:00 Backtesting, optimizations & over-fitting
00:31 Why Darwinex?
01:26 8 of the most important aspects of backtesting and optimizations
02:41 1 - A Trading Edge
03:59 2 - Statistical Significance and Sample Size
05:29 3 - The Dangers of Over-Fitting
06:08 4 - How to Avoid Over-Fitting
07:09 5 - Developing a Multi-Symbol EA (Expert Advisor)
08:45 6 - Optimization Design and Out-of-Sample Walk Forwards
09:37 7 - Interpreting Optimization Profiles
10:02 8 - Optimization Performance Metrics
11:03 Summary and Next Episodes

Content Disclaimer: Past performance is not a reliable indicator of future results. The contents of this video (and all other videos by the presenter) are for educational purposes only and are not to be construed as financial and/or investment advice.

[RightBackInRehab]

Just gone through your Algorithmic Backtesting and Optimization playlist for the second time. I can see this overview being incredibly handy for going straight to specific areas in the future. Thanks again for the work you put into these Martyn.

[Rizclinton]

thank you coach. the series are so helpful and I could skip all the guessing work

[leonjbr]

Good resume Martyn. I am really impatient to watch the series about Institutional-grade risk Management Techniques.

[tdb]

Very useful video guide Martyn, thanks for posting it!

[Martianx]

Very important checklist I never knew I needed. Thanks.

[davewilson]

Excellent material.

Regarding a comment you made - you said for a currency based system, you back-test "against a basket of 28 pairs". My question is, once you have the back-test data for the 28 currencies, how do you decide whether it is worth trading? So for instance if 10 of the currencies performed brilliantly but 18 were poor, do choose to go with the 10 good ones and ignore the 18? Alternatively if the 28 performed very average, but after analysis in a spreadsheet you found that a simple 'tweak' could make the performance very good, do you then make that tweak and go with the 28?

There are clearly other scenarios which could occur when you make your final selection, and obviously going live with the final 'combination' would be after a satisfactory walk-forward analysis.

It would be good to know how you make that final selection/elimination/tweak from the initial 28 currency basket.


Thank you for this work.

[subarkahsubarkah]

Hi Martyn, there are many parameters need to optimize (i.e parameter in trend filter indicator, opening/momentum indicator, closing indicator, etc). There will be a huge combination if we combine all parameters in a single step and therefore will consume huge RAM and also take long time to finish. Do you have any recommendation what steps I should do to do optimization in a well-structured step? For example, should I begin testing the parameter sets on opening indicator alone, or should I combine at least a set of opening indicator along with closing indicator? Or how? Thanks a lot, Martyn.. God bless you..

[rain]

Hi Darwinex, in you opinion what is a good amount of parameters a strategy should have for optimization?

[alessandraweberferreira]

Hi Martyn, do you know how to filter the paramenters before submiting the set of variables to perform the testing during the optimization? What I want to remove are conditions like BreakEven trigger < BreakEven profit... Do you have any video explaining how to do that by internal coding on EA?

[SupremeSkeptic]

Have you ever experienced getting 2 totally different looking graphs when backtesting an EA in MT5 using the exact same settings? One can look like a great strategy with a smooth slope, while the other a completely different jagged unprofitable graph.

Exact same settings and everything else. The bad looking graph would appear randomly.