I posted on this in “the New Grid” thread, but got nowhere, so I’m trying again:
If you sort The Grid by “Risk” (lowest to highest) the first page of 12 “lowest risk” systems is complete rubbish. There’s a couple of systems that might belong there, but not the other 10.
Anyone who is searching for a system using the Grid, adds filters then sorts the results by “Risk” (which is quite a natural thing to do) will soon conclude the results are nonsense. They will then ask themselves "can I trust any other information on this site?"
For the sake of the credibility of C2 I think “Risk” should be removed until the issues with it are sorted. Or just replace it with “% winning trades” from the old grid.
My version of the grid excludes options systems. Here’s what I see among the 1st page of 12 so-called “lowest risk” systems:
* 4 with negative annual return: -5% to -145%
* 7 with annual return under 10%
* only 3 with APD of over 0.4 (described elsewhere on C2 as a good APD and generally less risky than systems with low APDs)
* 7 with APDs near zero (0.02) or negative.
* 7 with negative Sharpe Ratios, and only 2 with Sharpes above 1.0
* 5 or 6 with impressive splurges of red ink on their thumbnail charts
* 1 (called JustATest) with a completely flat line on the thumbnail chart. The Sharpe of -30.5 and APD of -0.5 are some clues it shouldn’t be appearing!
I could probably go on, but I’m sure you can see that the results are nonsense. Do others get these results when they sort by “Risk” without filtering first?
There are two separate issues here.
First, there was indeed a problem with the way the “risk” column was calculated for some systems without enough data. This has now been addressed, and I think the fix has eliminated the most egregious of the problems Murray mentioned.
Second, I have renamed the column to more accurately describe what it is intended to portray. It is not meant to convey the risk of losing money. It is actually meant to convey the hidden risk component – the “iceberg” risk, I like to call it. I have already commented that the measurement is more or less the weighted sum of the following two stats: 1) the average intratrade drawdown as a percentage of account equity, and 2) the “risk of 20% ruin” as calculated in C2’s Monte Carlo simulations.
It’s not perfect, but I think it’s pretty good. The idea is to reveal risk that is not immediately obvious from results. I.E. The system might have been very lucky, and had several winning trades – all of which exhibited large drawdowns before the trades were closed. Or there might be a large discrepancy between the “actual” outcome and the probabilistic outcome (as determined by Monte Carlo sims).
So, there can indeed be systems that lose a lot of money, but whose hidden risk is not very high. That is, there is nothing hidden about the risk of these systems. What you see is what you get: the system has lost money, and it’s obvious for all to see.
This is a very different scenario than that of a system which has made money, but whose profitability is likely the residue of chance. Similarly, there may be systems that have lost money, but that could have lost a lot more money, and it was only due to chance that their results were just plain bad as opposed to frightening. These latter systems will score more highly on the “hidden risk” scale.
I have renamed the stat “hidden risk” to make all this (slightly) more clear.
MK
Thanks Matthew, that list now makes much more sense.
Personally, I have some major reservations re using Monte Carlo in that way, but then I’m no great fan of MC generally, other than as “academic interest” value.
Thanks for the quick fix.
