Are you sure about that? It seem to me it just depends on your percentile of 12P performance, compared to monkeys But in this case i am happy to be wrong.
I think that here is quite clear that recent returns weigh more than past losses.
Return and DD can be fooled by a lucky martingale or a manipulated inception.
You cannot turn a loser to a winner with luck, only with hard work.
Someone can explain me what the “return” in the performance pages means?
It’s not the same as the darwin return nor as the underlying strategy return, so what is?
Welcome to the forums.
The return is the same as the Darwin Return, but it is only refreshed once per day, while the Darwin return on many other places on the website is refreshed each minute, which is why you might see a difference.
The percentile table does not seem to include transaction costs. The 1st percentile should be more negative than the 99th percentile is positive, no? They both seem to be the reverse of each other (ie -65.3% vs +65.3%) which is impossible if we were to include transaction costs and swaps. Also the 50th percentile should not be exactly 0, it should slowly fade bellow zero return over time because of transactions. The monkeys are cheating!
You are right but the main problem with PF algo is the short lookback.
It mostly depends on the 3 Periods timeframe while the result of 3 periods is heavily influenced by luck.
If it were up to me I would consider 24 , 12 and 6 periods.
You don’t think the monkeys are useful at all? Even if they fix the weighting to recent data issue and the missing transaction cost issue and then also add more periods?
I agree, the weight should be on a larger sample size and larger dp data-sets should be available to select and view. But the monkeys in the chart are trading since the accounts beginning so with transaction costs properly included the chart might be useful I think.
With commission and spread missing from SUG (roughly 0.5% of the account per month is spent on commissions and I think roughly double that when we include spread.) in SUG. 1yr of monkey trading should have about 12% less in their returns after the 1st year is complete and since the monkeys appear to be trading from the accounts beginning no matter the dp selected… The monkeys not paying transaction cost does not seem to be a small problem either and in fact gets to be a larger and larger problem the older the account gets.
The monkey test makes sense but the sample of data is too small, I dont’ believe that you think that return of six months has some predictive power…
I will give you 2 examples of long trackrecords and the behaviour of Performance investable attibute:
A score with such volatility is useless to invest.
Performance (Pf) Investable Attribute: account with DARWIN vs. account without DARWIN
Account with DARWIN: real transaction costs included
We do take into account real transaction costs in the Performance (Pf) Investable Attribute when it comes to DARWINs since the Pf return comes from the real return of the DARWIN, therefore transaction costs included.
Account without DARWIN: transaction costs?
However, in the case of trading accounts with no DARWIN, the risk-adjusted return (10% monthly VaR) is just a simulation based on the underlying trading strategy.
Since such strategy’s return includes transaction costs, the simulation will calculate the risk-adjusted return with such costs although they won’t be as accurate as in the first case.
Hope I could dispel your doubts
OK but I’m not talking about the Pf score itself, nor am I talking about Darwins’ transaction costs. I’m talking about the percentile levels in the chart which represent the monkey trading. I’m talking about the Monkey trading transaction costs. The chart would be far more interesting if it DID include monkey transaction costs.
And btw, are you sure? I get that Darwins performance include the fees of the darwin but if the Pf algo includes the monkey performances excluding the monkey’s trading trasnaction costs then actually Darwins paying more fees are at a disadvantage since the monkeys that they are being compared to (or Pf calculated by) are not paying any fees.
Hi again @bendex,
The Pf is calculated differently depending on whether the strategy has a DARWIN or not.
If it does not have DARWIN, transaction costs are not taken into account.
In the case of DARWINs, and as a simplification due to a low monthly VaR, Pf is calculated assuming that the return curve behaves like a Gaussian bell. Therefore, we can draw the conclusion that in case a DARWIN gets a grade of 5, it means that it has neither gained nor lost anything so transaction costs are being taken into account indirectly.
That being said, and as you have mentioned, getting a 5 on Pf is more difficult as the DARWIN’s rotation increases.
Please, be reminded that we are currently working on improving our D-Score and some of our Investable Attributes.
May I wish you a nice evening
Yes that was my point thanks. This would be at least slightly unfair to Darwins who have a higher transaction cost per dperiod or per rotation though right?
High rotation DARWINs have more difficulties to become profitable due to transaction costs so it’ll be harder for them to get a 5 when compared to low rotation ones.
As I mentioned before, we are already working on ways to improve our metrics, so I just have to ask for a liiiittle bit of patience until we roll out the new stuff (FYI, it might still take a few months but it’ll be worth it)
Much shorter than I was imagining!
In light of another remake of the d-score coming up I’d like to say I agree with the above. And the weight being on the more recent of that data is also contributing to the volatility of this score which I think is too high.
Now that we have the New D-Score which focus heavily on the result/ performance (PF), I believe we should spend a lil’ bit more time here to discuss how the PF was calculated.
| did read the intro of PF that it is a peer-to-peer (Self-to-others comparison) on the current performance of all the monkeys on that particular month. So, for me it is kinda vague on how they compute the PF value without proper rigid (Self-to-self) formula.
By observing my own MUS darwin for the past few months. i’m guessing that PF isn’t just based on the result alone but also factor in the the RS (var) points. My RS had suffered huge drop during the Covid19 crush.
Despite the steep recovery of equity curve on my Underlying Strategy, my Darwin PF isn’t showing the same level of recovery. See below snapshots:
UL equity curve
Darwin equity curve
Hence, i draw my conclusion as follows:
PF is based on result + RS (Var)
PF is based on Darwin (Not UL strategy)
PF is peer to peer comparison.
If the above is true, then the new D-Score isn’t that bad since it also take RS (one of the vital IAs) into consideration. I might miss out something here. So, feel free to give your input.
p/s: MUS is experiencing the highest water mark currently when i stop focusing on those IA.
btw, my D-score went from ~50 to 37.9 after the new change.
The uderlying strategy recovered beacuse you incresed the risk from var 13% to 98% .
The darwin quote (and Pf score) reflects your real performance in pips.
If you lose 1000 pips with a size of 0.10 it wont work if you recover with 100 pips and a size of 1.00 .
Looking to trading journal you used to trade with a DLeverage of 10 , your recovery of April has been done with a DLeverage of 100 .