So, there is a relation between risk and duration. Sqrt(time) is a good evaluation. The result is obtain with classical equations (Vix or Black & Sholes reciprocal). If you dont like gaussian curve you may use power(0.6) rather than power(0.5), then you got the fat tail. It's a very rough way but better than nothing.
I definitely see the logic regarding the relationship, I just think the nature of the relationship is overstated. As far as I can tell (with my frustratingly small brain), the algorithm treats a multiple of Sqrt(time) the same as a multiple of leverage ie it treats a 0.1 lot size trade with a duration of 144 hours as having the same "market exposure" as a 0.6 lot size trade which lasts 4 hours.
Assuming a $500 account with 200:1 leverage, 50% forced close/wipe out: the first trade would require a 475 pip unidirectional move in 144 hours to wipe out the account (far from unheard of, but relatively uncommon), where as the second trade would only require a 58 pip move in 4 hours to wipe out (pretty regular occurrence).
Anyway the tl;dr is: I don't think this stuff should have anything to do with Experience.
Dale Pinkert makes some good trades using his moon planting guide
Yep, I'm open to being proved wrong, but I don't see how this system could be abused and would be a lot simpler and fairer.
My suggestion, should you wanted to increase the Ex attribute at a faster rate - keeping some hair "upstairs"
-, would be to prevent this one winner position from running for days on end and transform it into smaller positions so the duration of a position does not surpass a certain time threshold. You could close one trade -or even partially- and reopen it right away so positions will remain as homogenious as possible in terms of duration.
I know this is not the ideal scenario but, since Darwinex has not found a better way to measure experience coming from tons of different trading styles, this little tip will speed things up in the Ex IA.
This would be tough to code or have to be handled manually + create a whole heap of other problems ie massive divergence in R- consistency
As far as a better way to measure experience, I'm very interested in hearing your thoughts on simply having a minimum trade sample and time sample as suggested. I think 180-218 trade sample minimum with at least 12 months would cover all bases.
This is a very tough one, even my proposed system would still rank you relatively harshly, it would take ~4 years ... still better than 10!
“Experience” actually measures the "Statistical representativeness" of the data we have for a given strategy. In other words, given the sample we have, how accurate / precise are the rest of the attributes?
The problem is the algo is making arbitrary judgements about big chunks of the available data and throwing it out! As I pointed out earlier, an outlying trade is not at all representative, a representative trade would be modal
I'm leaving it there, sorry to anyone I didn't reply too - that was crazy
PS for context: OGL beat 19 of the top 20 Darwinia winners last month and didn't even make the top 48. I wasn't tracking where it actually ranked, but I have a feeling it would have been +100