Priority of variables in backtesting

Forums ProRealTime English forum General trading discussions Priority of variables in backtesting

This topic contains 11 replies, has 4 voices, and was last updated by avatar jebus89 1 week, 1 day ago.

View all attachments
Viewing 12 posts - 1 through 12 (of 12 total)
  • #73062

    Every back test has a detailed report and I would like guidance on the most important to the not so between:

    Max drawdown,

    Consecutive Losses in a row,

    Dollar return over period,

    Win rate percentage.

     

    #73064

    For me it would be in the order you list them above.

    I hate drawdown and dd is usually positively correlated with consecutive losses.

    I have Systems in Demo running that have around 85%  win rate but still lose overall … weird I know, but it’s all time to investigate / analyse and PRT Platform don’t make analysis easy!

     

    #73065

    They are all equally important.

    Big drawdown will most likely kill your account unless you are Warren Buffet

    Lots of losses in a row will potentially kill your account or make you turn the strategy off.

    Low return over the period means you have to risk more to make anything sensible plus the lower the gain per trade the closer you are to the cliff edge of failure.

    Win rate percentage – low is OK if the wins are massive but most prefer a higher win rate. Psychologically it is easier to trade and more wins usually means less draw down.

    IMHO

     

    #73145

    When ur backtesting u want a smooth EQ curve and pretty much the same avtg win pr year every year if possible. If your backtest have 1-2 REALLY GOOD YEARS and 5 very much less profitable years, then uve probably curved it for those 2 years.

    if u tho have 10 years all look around the same avg gain pr year then thats alot better. u can also view the smoothness of the EQ curve. u dont with high spikes here and there where backtest is perfect, and then crap other places..

    avg gain pr trade is very important cus its gonna tell u something about how much u can afford to loose on each trade. if u make 15$ pr trade then u shouldnt worry too much about spread and slippage. if u make 2$ avg gain pr trade, then u rprobably gonna loose those 2 to spread and slippage (unless uve calculated backtest with correct avg spread + add something for the slippage..

     

    Max DD is all about getting it under a limit that ur comfortable with. I wouldnt run something that have 3000€ drawdown trading a 1€ contract. but thats me.. dosnt mean the system cant be profitable, just means my limit is a long time b4 reaching 3K drawdown..

     

    Max loss cons. is gonna tell u how much pain u gotta take. If u see 22 max loss in a row, then u have to sit through 22 painstaking losses, each loss making it harder and harder to keep running the system… (u probably need a better filter lol)

     

    Winrate is of course important ass well, but u gotta be realistic about ur winrate in my opinion.. If ur trying to catch every trend going up, then ur gonna probably look at 35% -> 50% winrate. If ur trying to mean revert everytime price is extreme volatile then u should hopefully be look at 60-80% winrate.. then ofc u can add trailing stops etc to make some strategies better.

     

    When it comes down to “what should i focus on”? Everything hehe! If all of the above is looking good u should see steady profits each year..

     

     

    Edit: i can add that when u backtest u rly wanna scroll through the top 50 or smthn.. cus even tho #1 might have the most profits, if u move down to say #12 u might miss out on 200€ in a totalt of 10 years or whatever, but ur EQ curve can look so much better and even.

    1 user thanked author for this post.
    #73146

    Edit: i can add that when u backtest u rly wanna scroll through the top 50 or smthn.. cus even tho #1 might have the most profits, if u move down to say #12 u might miss out on 200€ in a totalt of 10 years or whatever, but ur EQ curve can look so much better and even.

    Which is why it would also be great if the results window included a column for maximum draw down and lots of other helpful information to sort by.

    In addition to your comments I would add that the number of trades is also important. Too few trade and the data sample you have to work with is too small to have much confidence in it. Evenly spaced positions throughout the test period is a good sign too – you don’t want one trade a year and then twelve in the last year for example.

    Time in the market is also important as well as average and maximum trade length of time. Time in the market is time with risk on and just one trade with a very long time in the market can eat into profits with fees – plus could you actually sit there for six months watching a trade go up and down and sucking fees out of your account or rolling over and costing you spread each time when most of the other trades were in and out in a day?

    1 user thanked author for this post.
    #73151

    sit there for six months watching a trade go up and down

    Yeah and you may even get an ulcer, cancer, a coronary or a divorce due to stress! 🙁

    I look for no of bars in a trade to be broadly related to the TimeFrame of the strategy else I consider the System is just waiting and hoping for a passing favourable wind to blow it back  into profit.

    I have no fixed rules, but no of bars would be related to the natural cycles of the instrument traded. So for Dax on a 1 min TF maybe 20 mins to 1 hour (20 to 60 bars) for example during the day, with maybe longer during the night in a slow moving market etc.

    If a strategy is on a 1 min TF is in a trade for > 4 hours then it probably isn’t suited to the 1 min TF??

    Just a few thoughts and generalisations I look for when backtesting
    GraHal

    #73156

    Thank you for all the replies.  How many bars of time should a trade be in on a 1 hr, 4hr and daily TF before its overstayed its welcome?

    On another note not part of this topic but if you optimize and have back tested for 10 yrs why do I still need to go to demo before I go live?

    #73160

    why do I still need to go to demo before I go live?

    Because otherwise you will never truly know how curve fitted the strategy is. Every strategy is fitted to some extent but true OOS forward testing is the ultimate test – it is up to you whether you decide to use real money or pretend money. I believe you have tried the real money approach and found out the hard way that pretend money may have been a wiser choice! 🙂

    #73165

    True and point taken

    #73179

    How many bars of time should a trade be in on a 1 hr, 4hr and daily TF before its overstayed its welcome?

    I wouldn’t like to make a declaration on rules, but you can see if a curve is going through lots of major highs and major lows and it is not exiting I ask myself … wtf!? 🙂

    I do a lot of manual trading so I’m all about market structure (major highs / lows, support / resistance levels, Fib retrace etc etc) so when a System ignores market structure (rides the cycles etc) I get nervous! 🙂

    How about you post an equity curve of one of your Systems and I / we would give an opinion? It would be interesting to read different responses from folks?

     

    #73182

    When I have a decent system I will post it. I am still in early days

    #73188

    Thank you for all the replies. How many bars of time should a trade be in on a 1 hr, 4hr and daily TF before its overstayed its welcome?

    On another note not part of this topic but if you optimize and have back tested for 10 yrs why do I still need to go to demo before I go live?

    How many bars…? All depends on what ur trying to do.. If ur trying to rode a 1 month long trend then ur gonna need some bars. if u wanna mean revert ur probably only gonna need 2-10 bars… who knows 🙂 all depends on what ur trying to do, there is “right answer” to this..

    Why would u put in demo…? Ive had many strategies that has looked good but then after 2 months of demo, even if market has gone up, it has just fucked itself with like 10% winrate and constant loss… shit happens… dont rush it, this is a marathon not a sprint.

Viewing 12 posts - 1 through 12 (of 12 total)

Create your free account now and post your request to benefit from the help of the community
Register or