P/L Unit: Useless & Misleading

… but mostly just useless, because its so misleading.

1) I’m not concerned about this value and how it affects my system here on C2.

2) I am always on the lookout for systems worthy of my trading capital on C2. An accurate P/L unit measurement is necessary as part of my analysis.

That said, whatever algorithim being used on C2 makes no sense. I can’t make hide nor hair of it after six months of observation. Whatever was trying to be accomplished with this number has failed.

As an extreme example: Breakout ES. This program has traded 286 contracts. It’s gross profit is $36,624. It trades on one symbol only, so that isn’t an issue. Its actual P/L per unit is $128.06. The C2 P/L Unit is MINUS (2.27). (!!!)

Other systems have exagerated P/L Unit to the upside, although not equal to the reverse of this example.

Any system I look at I have to manually copy/paste each 25-record page, add the contracts, then divide them by the profit. This is just silly.

Additionally, it seems to affect the “actual slippage” numbers. This is a great addition to the sight, but it loses all value when the numbers don’t add up. Specifically, no system I have looked at has consistent/comparable numbers under the “P/L Unit after real-life slippage” and “Cumu $ … after real-life sllippage”.

In fact, they very often give exactly opposite pictures of “real-liffe” slippage. I would be happy to give specific examples, but just looking at any “best systems” page should give plenty.

Matthew, I hope this gets fixed/changed.

As I have pointed out earlier, in a closed thread, the P/L per unit at C2 seems to be computed by first computing P/L per unit for each trade, and then averaging it across trades. The outcome of your (imho more adequate) formula is also affected by the relative position sizes, and so it can be very different from the C2 outcome. Something that was denied by Sam and Ross for about 100 posts in that closed thread, so I won’t say another word about it. Note however that, as Sam correctly pointed out, the difference can also be caused by open positions.

Nothing I am talking about here is a result of open positions.

I am also specifically addressing systems that only trade one instrument, to keep things simple.

Whatever this measurement was designed to represent, it is simply not doing so.

No one I’ve communicated with, or heard from, understands it whatsoever.

Hi Dustin

The P/L per unit for each trade gets calculated and then averaged. So, if a system a have a couple of big winners with lots of contracts and a bunch of losers with much less contracts, but larger than the P/L per unit for the winners, then the P/L per unit can be negative.

Reading this last sentence, I am not even sure I will understand that myself when I read it tomorrow, so let me try to explain with an example.

Say you have two trades as follows:

10 Contracts, $1000 profit

1 Contract, $200 loss.

Total profit = $1000 - $200 = $800.

P/L per unit for first trade is $1000/10 = $100.

P/L per unit for 2nd trade is $200/1 = -$200.

Average P/L per Unit = ($100 - $200)/2 = -$50.

The other way is what you did:

($1000 - $200)/11 = $72.72.

Which one is more correct? Probably both. But my preference is for the way it is currently calculated. If the above system has a subscriber who trade only 1 contract for each signal, his P/L would indeed be -$50. In that sense, I prefer the C2 calculation.

I didn’t look at ES Breakout system you reference below, but I would bet if the system traded only contract for each signal, it would have been a net loser overall.

I know some people will argue that some signals can have higher probability than others, so a system might trade more contracts on those and less contracts on the lower probability and as such P/L per unit should be calculated by Total Profit/Total Contracts. And that probably is a valid arguement and probably better left for another forum topic.

My preference still is the way C2 calculates it as this is what the 1 lot, or fixed lot trader can expect. I don’t think it is useless and misleading.


- Fanus

"Whatever this measurement was designed to represent, it is simply not doing so."

I tend to agree. I’ve said it a 100 times. But do you understand the difference between the two formulas?

If there are 1 lot systems where you see these strange results, than I would say there is an error in their calculation. However I just looked at Breakout ES which you refer to and on the fist page, the trade sizes differs from 4 to 24, which would explain the negative P/L per Unit based on C2’s calculation.

- Fanus

Dustin, please forget my previous post. Fanus has already explained what I meant. And I agree with him that the current formula is useful for someone who trades with 1 contract all the time (or any other fixed size). The other formula would usually be more useful for me though.

Hi Jules

Seems like we are typing at the same time. :slight_smile:

I don’t think this would be very difficult for Matthew to calculate both and display it on the statistic page and I can see use for both. I can also see how two P/L per Unit values on the statistics page can be confusing and will probably raise a lot questions.

- Fanus

Yes, we were :-). And you was more awake than I, because I initially forgot the case of someone who trades with 1 contract, although you have pointed out that before. But it is 7:45 AM here…

I agree that displaying both would be the best.

There it is, thank you.

It turns out this statistic is not complicated after all – I assumed 1 contract per trade for Breakout ES, and the figure matches exactly.

Nothing is averaged, etc, as many were guessing.

This is actually then a very useful metric, but no one knows what it is! Several threads on this metric over the past few months, and nothing but confusion as to what it is and what it’s supposed to be.

It needs to be renamed (it is specifically NOT “P/L Per Unit”, but rather P/L One Unit, and have one of those little “Whats this?” icons next to it.

"Nothing is averaged, etc, as many were guessing."

Huh? Fanus averaged across trades in his example.

I agree with your renaming suggestion, although I can imagine that the present name is convention.

His explanation was correct then, although by his own admission confusing (I focused on his description rather than his math).

Now that I know specifically what the metric is (after six months), why the heck is it only accounting for 1/2 the stated commissions, and 1/2 the observed “real life slippage”? Matthew, I’ve asked this question on two other occassions with no reponse or changes.

Sure, it’s easy to do that bit of math in one’s head, but to my mind it really degrades the effectiveness of C2’s presentation when the numbers don’t add up correctly.

Dustin: What do you mean when you ask:

"why the heck [is the Avg P/L per unit stat] only accounting for 1/2 the stated commissions, and 1/2 the observed ‘real life slippage’?"

The P/L per unit stat has nothing to do with either commissions nor slippage. It is very simply a metric designed to tell you what sort of average profit you can expect per traded unit on average.

Can you explain your question a bit, and then I will try to answer it? (Sorry for the lack of clarity.)

Only $2.50 per trade is subtracted in the "after typical commissions" field, rather than $5.00.

The slippage number, while presented as a figure "per side" in the mouseover text, is only accounted for once rather than twice (once for each side of the trade).

Sorry. I still don’t understand. It sounds like you have questions/problems with the way the Average Commission and Slippage stats are calculated (which I would be glad to discuss further) but let me say again that both these statistics have no relation to the Average P/L per unit number.

After more examination, I can’t imagine a use for this metric.

After figuring it out, I assumed it was used to measure the effects of trading a single unit (or unit size) rather than trading variable unit amounts.

It doesn’t measure this however. When I add up the P/L for trading a single contract of Breakout ES, it’s not even close to this measurement.

How is this number in any way useful in evaluating a system?

I don’t have any problem with the way slippage or commissions are calculated, not at all.

You use $5.00 as typical commissions for the Cumulative $ figure. You use $2.50 as a typical commissions for the P/L figure. Is the typical commission $2.50 or $5.00? It can’t be both.

Mouseover text indicates calculated slippage is per side. In other words, if slippage is $2.00 average per side, it is $4.00 per trade (a trade has two sides). You are only subtracting the $2.00 rather than the appropriate $4.00.

"It doesn’t measure this however. When I add up the P/L for trading a single contract of Breakout ES, it’s not even close to this measurement."

When I do the computations for Breakout ES according to the formula that Fanus and I tried to explain, then I arrive at ($2.27), which is exactly what C2 displays (before commissions).

If the commission is $2.50 per side of the trade (i.e. $2.50 for buying and $2.50 for selling) then after commission the number should be ($7.27), while C2 displays ($4.77). So either C2 forgets to subtract the commission twice, or it assumes $1.25 commission per side.

About the calculation, at first glance I thought it was useful to describe profits trading a single contract as well.

However, that’s not the case. If you trade one contract you get in along with the first contract, and out when the last contract closes. So for example, your profit trading this system with a single contract on the first trade would have been $400.00. Instead, this P/L per unit measurement counts your proft as $355.00 for that contract.

For an even more exaggerated example of this, look at the most recent trade for my system. A single-contract trader would have made $475 on his single contract trade, but P/L per unit assumes he made $33.00 on his single contract.

It just doesn’t work.

I see what you mean. It is a problem for multi-legged trades. Indeed, it doesn’t necessarily say what a 1-contract trader would get. For that purpose it should (also) be computed differently.