Among the statistics, the Compound annual % return should be changed back to the original Linear (or just annual) % return. I know, the compounding makes it bigger but:
1. Not every system compounds gains.
2. The compounding rate can be different for everybody, so whatever Matt is using is very arbitrary.
3. Compounding also makes for more risk beside bigger returns, so let’s say a 20% previous DD after compunding can make a 40% DD 2-3 months later.
Thus let’s use the original linear method. If a system is up 25% in 3 months, the annual should show 100% projected for the whole year…
Yeah, I’m not getting this one either, MK.
I been up 5-15k (approx) and “compound annual” kicked in nicely - like 25% and then the last few days I’m up 7-10k - but my c.a. % is ~5%???
Very confusing.
Again, please take time to help what has been outlined (like my case slippage/commissions/compound annual). Thank you.
Gilbert
[LINKSYSTEM_26557785]
I am against that suggestion. Please note that both versions (compounded and uncompounded) are reported in the advanced statistics. So if you really want the other version, you can find it there. But for the basic statistics table, I believe the compounded version is more appropriate, for these reasons:
1) I believe the large majority of systems is compounding.
2) It is not true that the compounded version is always larger than the linear version. For profitable systems older than one year, the compounded version is smaller than the linear version. E.g. see the advanced statistics of extreme-os, hawk-fx, etc.
3) For this reason people previously complained that the linear version was misleading, e.g. in the case of hawk-fx.
4) Note that holding a profitable position for more than 1 day is actually also a form of compounding.
I think that the situation is much worse with the linear version. Profitable systems that are older than one year are most likely to be considered by newbies. But exactly in these cases the linear version would be misleading and too high. Moreover, the situation will become worse and worse when the system grows older. For systems much younger than a year, on the other hand, everyone can understand that the extrapolation to annual results is rather speculative.
PS I don’t understand your point 3.
Nevertheless you gave a pretty long answer. 
Anyway, if a system is linear and it had a 20% DD let’s say twice, then we can expect that in the future. If the system is compounding, that very same 20% DD could become easily 30-40% because compounding works both ways. So the expected/projected DD is missleading…
Well, I have to say, I never clicked on the Advanced stats before and I think it is a bit of an overkill, but sure the linear annual % is there. So thanks for the suggestion, now let’s adress your points:
1. Most systems are LOSING, so they are definiatelly NOT compounding. There is nothing to compound. Period…
2. Here the point is, which is the more realistic and which shows the valid expectations? In linear annualization it doesn’t matter if the system is 6 months or 2 years old, both ways it shows the average what can be expected after 1 year or 2…
3. There are very few systems like Hawk-fx, so the major stats should carry what the majority of the system do. The few systems like that can look it up among the advanced stats.
4. Simply not true…
The rest: I don’t see the problem with using the linear method with older systems. As long as the system is going up with the same ratio as in the first year, the % should remain the same. So what is the big deal? If the system stops going up, the annual % should fall and it will, thus correctly showing what can be expected…
I think it is just an advertising gimmick, quite a few people fail to notice that currently it is a compounding %.
Also I am just curious how the compounding ratio is calculated. Is it a fixed number or does it grow with the system itself?
Jules, you are so wrong on so many points, it is not even funny. I took a look at the mentioned systems:
Extreme-os is NOT a compounding system. It has one of the straightest equity curve, with compounding you get an exponential curve. The system is actually being shortchanged, because it has an actual arithmetic return of 262%, and the compounded rate showing is only 118%, so it is bad advertisement for the system.
Hawk-FX: again, compounding is showing a 86% annual return when the arithmetic real return of the system is about 160%, again, bad advertisement for the system…
It could be that there is a problem in the C2 calculations or maybe in mine, but both systems were better off having the arithmetic annual shown…
I agree with Jules, MK please do not change it. If people need to see the linear version they can look in advanced stats, but in my opinion showing the compounded annual on the front page makes much more sense.
Perhaps as a compromise we could have another explanation (? icon) alongside to say exactly what it means so people are aware. And while you’re at it perhaps you could add one for the p/l per unit so that it makes sense to those unaware as I’m sure there are others like me that would prefer to see it weighted.
Sorry, I think you’re wrong on all points.
1. If a system is losing, it cannot invest the lost capital, which means that it obligated to adjust its trade sizes to the available capital left, which is a form of compounding. Not compounding would mean that it can start again and again as if it has $100K (or whatever the start capital was). This is not possible at C2. So for losing systems the compounding formula is more appropriate. E.g. if the system loses 1% at its first day then the linear formula would predict that it will lose about 250% in a year, which is obviously impossible. Anyway, for losing systems the discussion is not very relevant, because no one will consider to subscribe to such a system.
2. No, for profitable systems older than one year the linear formula will lead to an overestimation of what the annual returns of the system, unless the system is not compounding (i.e. uses fixed trade sizes throughout its history).
3. I think the majority of the systems adjust their trade size to the available capital, which means that the compounding formula must be used.
4. On the contrary, it is true. E.g. if the system buys $100K of QQQQ on day 1, and it keeps that position for a year, this is compounding. Not compounding would mean that it adjusts its position to the value of $100K every day. I.e. when the position becomes worth $101K it should reduce the open position to $100K. Nobody is doing that.
“The rest”: Yes, if the account value of the system continues to grow with the same rate as in the first year then the reported % should remain the same, and NO, the linear formula is not doing that. The linear formula will yield an ever increasing percentage, even though the system is growing at a fixed rate. That is the problem.
With respect to the DD’s: Me saying that I don’t understand your point was a polite way of saying that I think that you don’t understand it Suppose the system has $200K and it loses $20K, then the DD will be 10%. Later, if the system has $400K and it loses $20K again, then this new DD will be 5%, while the first DD will still be reported as a 10% DD. This is consistent with compounding, not with the linear formula.
How the compounding annualized return is computed: You can look it up in the guide of the advanced statistics, and in many other websites.
With respect to extreme-os: You are wrong, that system is compounding. The linear curve is the result of a gradually detoriating performance. If you go to the system you will see that it has currently more than a million $ in invested in long equity, while it started with $100K. This implies that the system is compounding, i.e. the trade sizes are adjusted to the current account value. A non-compounding system starting with $100K could have invested at most $200K in equity positions that are held overnight. For what it’s worth, before I subscribed to that system I too believed for a while that this system was not compounding, for the same reason. But it simply isn’t true.
I started to care less about the issue, so at this point just a few points:
You are right about extreme-os, it is compounding and that is one bad system. The interesting thing is that the compounding makes up exactly for the deteriorating performance.
Now you didn’t address how come that compounding % was actually showing LESS than the linear % in case of 2 winning systems.
>Yes, if the account value of the system continues to grow with the same rate as in the first year then the reported % should remain the same, and NO, the linear formula is not doing that.
I don’t know why. It should be (unless I am mistaken) a simple division, gains divided by years.
> Now you didn’t address how come that compounding % was actually showing LESS than the linear % in case of 2 winning systems.
This will always happen if the system is profitable and older than one year. What you see is the rule, not the exception. (The rule for profitable systems older than one year, that is). I don’t know at what level you desire an explanation:
- If an example is enough for you: Suppose the system gains 50% every year and it starts with $100K. Then after year 1 it has $150K, and after year 2 it has again gained 50% from that, so the total account value will be $225K. The compounded annual return is then ($225K / $100K) ^(1/2) - 1 = 0.5 = 50%; exactly what it should be. But the linear version yields ($225K / $100K - 1) / 2 = 0.625 = 62.5%. This is too much. You will not gain 62.5% every year, you will gain 50% every year.
- If you want a proof: Suppose V is the last account value, V0 is the start account value, and t is the time passed, measured in years. Let x = V / V0. Then the compounded annual return is x^(1/t) - 1 and the linear version is (x - 1) / t. If x > 1 and t > 1 then the latter is larger than the former. This follows from the fact that they are equal at t = 1 and considering the first and second derivative with respect to t.
With respect to extreme-os I must add this to what I said previously: If you look at the chart with logarithmic axis then you see that this is almost linear too. So the phenomenon that the other curve looks linear is not only caused by detoriating performance, but also by the fact that an exponential curve is hard to distinguish from a linear curve if the exponent is small and you see only a relative small part of the curve.
>But the linear version yields ($225K / $100K - 1) / 2 = 0.625 = 62.5%. This is too much.
See my version of the linear math is this: gain divided by years, so system is up 100% in 2 years equals 50% annual just like it is and not 62.5% what you got.
But if it gains 50% each year then it after two years it will be up 125%, not 100%, if the system is compounding. So "your version of linear math" will yield 62.5% too. Also, you referred to the way in which it was computed previously at C2, and that formula would yield 62.5% in this example.
