From observing the stats, my behavior as a subscriber and the behavior of strategy leaders, I would argue that the way C2 handles, e.g., winning percentage is wrong. The reason: C2 takes all individual positions together and only looks for the bottom line. Thus, strategy leaders are incentivized to average down as this way they can still have a high winning percentage. If the winning percentage would look at the individual posiitons, this would be corrected as then some of the positions would be bad, while others may be good, giving a better picture of the performance.
Alternatively, it would be good to have this as additional statistic (winning positions, calculated on a LIFO-principle)
Winning percentage is a very misunderstood statistic. Newbies tend to think it’s more important than it is and misinterpret what it does mean. By itself it’s almost meaningless. A conservative system that cuts losers short might have several small losers to every successful trade, but can still be profitable if the successful trades gain more than the combined losers. Alternatively a reckless system might have 99% winning trades and be still be catastrophically unprofitable if the 1% of losers lose big.
When I’m evaluating systems I see a high winning percentage as a red flag. High %win systems tend to hide risk. Nobody can keep a very high winning percentage unless they are averaging down or holding losing trades until they turn around–both strategies I want to avoid. On the other hand I’m extra interested in low %win systems that still manage to be profitable as they tend to be lower risk.
IMO unsophisticated investors that seek out high %win systems are ironically doing the opposite of what they intend–they are filtering out safer systems for systems with hidden risks that will one day rise up and bite them.
9 Likes
David, this is all correct, what you say, but not my point. Also system developers (most) fall into this trap, either unconsciously or in order to lure newbies. And this could be easily fixed by computing the states really for each individual trade, instead of only when a position is completely closed.
As a side-effect it would resolve a problem some develops complain about that if they create a position and then increase and reduce it repeatedly C2 shows their system as not having separate trades and hence not as very active…
1 Like
See FiFO vs LiFO. C2 is using LIFO so unless they are going to introduce also FIFO, statistic like the Nr. of trades or winning percentage are meaningless to evaluate multi-strategies type program listed on C2. Many of us have already asked many times (recently and in the past) to introduce this important feature… Unfortunately It does not seem a priority for C2.
MechStrat - I do not think they are using FIFO or LIFO. It seems they only count total closed positions. FIFO or LIFO, both would be an improvement.
would having a winning percentage per day be useful?
@KlausS Could you please put some details on how the # of trades and winning % are calculated using your approach? I am not familiar with it, never seen it in different trading software, and always used conventional approach like C2.
For example the transaction sequence is as follows:
Buy 100 shares x 100$
Buy 100 shares x 98$
Buy 50 shares x 95$
Buy 250 shares x 90$
Sell 200 shares x 100$
Sell 200 shares x 99$
Sell 100 shares x 101$
Based on conventional approach it will be 1 trade with 500 shares with P/L of +2850$.
@AndreyBlinkov I am not sure about other trading software. But here is how it would work for FIFO:
Wenn you sell 200 x 100$, you actually sell 100x100 and 100x98,
so these are two closed positions, the first has P/L=0, the second has P/L=$200.
Then you close again 200x99, these are 50x95 : P/L=504; and 150 x 90: P/L=1509
the last 100 are 100 bought at 90 and sold at 101 = P/L= 100*11
This is actually how it works for tax purposes in Germany (to my knowledge).
Actually, I think it makes more sense, because if you sell s.th. you are trading, aren’t you? So, why shouldn’t this not be called a trade?
HI Andrey,
Your diagnosis is correct and following your trades sequence, Fifo would provide an outcome of 4 trades.
Perhaps 2 examples will clarify better the weakness/strength of these 2 approaches:
+1 +1-1 +1-1 +1-1 +1-1 +1-1 -1
C2 = 1 trade and report a position long 6 contracts
FIFO = 6 trades each of 1 contract.
Conversely:
+5 -1-1-1-1-1
C2= 1 trade and report a position long 5 contracts
FIFO= 5 trades of 1 contract each
Note that It depends of the type of the study and system being reviewed that make the difference. There is not a superior method. But would be nice have available both methods.
Thank you! Now it is clear for me.
Am I correct that in case of the single position closing transaction (like buy 1, buy 1, buy 1, sell 3) it will be only one trade?
In 2nd case FIFO gives really bad results, position size (and related risk) is under-estimated significantly. Looks like C2 approach is conservative. Which is good for subscribers. 
Regarding the trades, because you have 3 different openings, you have three transactions.
Regarding position size:never said everything should be computed that way. For example what you mean is actually max position size, I.E. how big are simultaneously open positions. These would be 3in the example, while the individual trade sizes very from 1-3. But both are actually without any interest. For example, how compares a strategy that does buy a, buy b to a strategy buying 2 a. If a and have similar risk both strategies should be seen similar, but first would​have position size 1 for c2 while the second has 2.
I got it now, thank you for explanation. For myself I don’t see any benefits from additional numbers, but maybe C2 will consider it. Also I am more interested in max position size rather then single entries/exits size. Max position size defines risk taken by opening the position.
I think it is not as easy to game as the current system. Right now many add to loosing positions to improve their stats. For such systems winn%. would be more realistic with this approach. While the C2 approach make those look good, this would be less so with this approach as some of their positions would still be losing. Thus it would even disincentivize such behavior.
Looking at specific trades / entries / exits makes it easy to identify martingale practices. If a large position is a result of trades losing money, this is an indication of both bad timing (of initial entry) and bad risk management. Of course, the maximum position size / exposure is also very important.
