What to look for in a Trade Explorer
Somebody suggested a post on this topic, and I couldn’t resist giving it a shot. My comments are based on a hypothetical scenario: what would I look for in a TE, if I was considering having the TE’s creator trade my money on my behalf.
First, it’s important to define one’s desired performance agenda. For me, it’s steady long-term account growth with tight control on risk, and my priorities are set accordingly. (For others, it might be to start with a $100 account, the loss of which they can easily shrug off, and use maximum available leverage in the hope of turning it into $100,000 as quickly as possible. For still others, it might be the guaranteed recovery offered by a martingale variant, because they are willing to wear the risk of sudden and unexpected loss of the entire account, at any future point. In short, forex is an open-ended vehicle that allows plenty of different speculative strokes for different folks).
The numbered points correlate with the numbers in my attached screenshot.
1. Statistical validity: This is the first thing that I look for. Given my long-term agenda, I like to see an absolute minimum of a 12 month, 500+ trade, track record, on a live account, and preferably a lot more. Anything significantly shorter than this is unlikely to be indicative of future performance, i.e. could be written off as “beginner’s (good or bad) luck”. Time is more important than number of trades, in that I believe it's important to demonstrate survival across as many different market conditions as possible. (However, offsetting this a little, I also understand that markets can 'evolve' over time [some possible reasons: economic, technological, trader belief/behavioral shifts], which means that distantly historical performance runs the risk of being unrepresentative of current market conditions).
Statistical longevity is important for two other reasons:
(i) many systems can be propped up for a long time by MM that is ultimately unsustainable, e.g. martingale variants; or otherwise rely on some kind of mean reversion, i.e. counting on open losses to recover, that will eventually be tested by 'black swan' or 'fat tail' moves; and
(ii) it is easier than many folk realise to be 'fooled by randomness', as is explained very well here.
Of course I realize that FF’s TE hasn’t yet been around for long enough to allow for track records of a decent size.
2. Profit factor: I once read an article in which a famous trader (it was Paul Tudor Jones, if I recall correctly) said that he could count on the fingers of one hand, the number of traders he knew personally that, over their entire career, had been able to maintain a PF>2. Hence, even allowing for the fact that retail traders have the advantage of agility, whenever I see a PF>2, I think something like “well either
(a) this guy is well on his way to becoming a multi-millionaire, or
(b) he’s employing some unsustainable MM, or
(c) if he doesn’t meet the criteria in point 1, then he’s been lucky so far”.
Any EA with a PF>2 almost certainly fits into category (b) or (c). If it seems too good to be true.......
IMO a better PF is obtained by dividing the profit pips by the loss pips. That cuts MM (point (b)) out of the equation. As I’ve posted many times before, MM can ultimately never rescue a trader/system whose entries/exits fail to attain positive expectancy. Trade expectancy determines whether your equity curve points upward or downward, and MM (position sizing) determines the steepness of its gradient. Increase position size and you increase return and risk in like proportion. In this post I illustrate how it’s possible to turn a 30% per annum return into 900%, by merely using MM and increasing trade & compounding frequency. If a trader/system has attained a pip-based PF>1.5 over more than 2 years, and 1,000+ trades, then that’s a symptom of a very proficient trader, IMO.
One very important point: PF must include open trades, i.e. unrealized P/L. A number of traders, instead of applying a formal loss exit strategy, leave trades open until they return to profit (or, even worse, average their losers). Unrealized gains and losses form a very real part of the available balance in a trading account.
3. Maximal drawdown: this is essentially the product of MM and risk management. If it has been allowed to reach > 30%, then the trader/system is sizing too aggressively for my personal liking. But, having said that, everybody’s appetite for risk is different. Remember that, if you’re using fixed fractional sizing (compounding your equity growth, by risking X% per trade), when your account is Y% in drawdown, it requires a gain of (100 x Y)/(100 – Y) percent to return to breakeven. For example, if your account is 20% in drawdown, it requires a (100x20)/(100–20) = a 25% gain on the remaining capital to return to BE. Hence the more you let drawdown run away from you, the more irretrievable it becomes.
In addition to this problem, risk of ruin increases geometrically when position size is increased. For example, if you double your trade risk from (say) 2% to 4%, your risk of ruin — however small it might be — is nonetheless more than doubled.
In general, risking the textbook 1%–2% per trade should keep you out of trouble; at least it will afford you some time to study your system performance, and if necessary, revise and overhaul your system, before drawdown gets completely out of control. However, if the trader/system has an exceptional win rate, then there may be reason to justify an increase beyond 2%. The table attached to this post gives the probability of attaining X consecutive losses over the course of 50 trades, for a given win rate. However, keep in mind that the table assumes 1:1 RR trades, and also does not take into account losing sequences that contain occasional wins, e.g. L– L– L– L– L– W– L– L– L– L has only 5 consecutive losses, but the losses exceed the wins by 8. Again, what constitutes acceptable or unacceptable drawdown is a personal decision, and it may be worthwhile to run some kind of Monte Carlo-type analysis (assuming that you have enough historical performance data) to gain some kind of feel for your optimal position size.
One final point: some traders don’t have all of their trading capital deposited with the broker, hence a X% drawdown in the account may be significantly less than a X% drawdown overall. (Obviously it’s only necessary to keep enough money in the account to avoid a margin call; the rest may be deployed safely elsewhere).
4. Lots per trade: I look at a breakdown of the individual trades, both for the use of stoplosses (alas, FF’s TE doesn’t show these), and unduly large variations in the lot sizes used. Large variations in size is a big red flag to me. Formal stoplosses aren’t essential, but if they’re not there, it’s impossible to determine what kind of loss exit strategy (if any) the trader is employing. If there are signs of averaging down, or (even worse) doubling down, then that’s the biggest red flag of all, IMO. Obvious signs of ‘hedges’, baskets and/or recovery/rescue trading are also a big turn-off for me, as they provide no real edge, and suggest that the trader lacks a basic understanding of trading math.
I work on a basic rule of thumb of 1 full lot per $25,000 in the account (scaled proportionally according to the account size). That equates to approximately 2% risk with a 50 pip stoploss. If position sizes are significantly larger than this, then I start to question the level of risk that's being taken. (See the XLS/image below: for example, at 2% risk and a 50 pip SL, you may trade 0.04 lots for every $1,000 in your account).
I also look at the largest losses in the account, percent-wise. Again, each loss should be no greater than 2% of the account balance, unless there are valid reasons for taking greater risk (e.g. an exceptionally high win rate).
5. Average win size divided by average loss size, preferably in pips. Most profitable traders/systems end up with a career RR>1, i.e. they win more on winning trades than they lose on losing trades. Hence a value > 1 is a healthy, but not necessarily essential.
I also compare the loss sizes (percent) with the win sizes (percent). A consistent highest %loss usually means that formal SLs are being used, which is good risk management IMO. And again, the loss sizes should, on average, be less than the win sizes, which is an indication of a healthy RR.
Finally, I check the winning trades (in pips) against the trade time. If there are many small wins (e.g. < 5 pips) for trades that have been open for several hours, this suggests that the trader might be playing some kind of recovery strategy, i.e. leaving trades in loss until they return to breakeven, a dangerous strategy that lacks 'cut losses' risk management. I'll open up the charts of such trades, to see how far they were allowed to proceed into drawdown; if the worst case is several times the size of the average winning trade (in pips), then that's another red flag.
For me, those are the “big five” criteria. The Sharpe and Sortino ratios give a measure of the consistency of monthly returns (higher values indicate greater consistency), but these are secondary, only after one has attained some evidence of positive expectancy.
Balances and percentages are much less important to me. These are the figures that attract newcomers to forex, but they can be easily skewed, not only by MM, but also by deposits and withdrawals from the account. With leverage and some luck, almost anything is possible — in the short term.
I’ll add to this post when I get some more time. But that’s enough for starters. I expect that it will provoke some interesting comment.
UPDATE - Feb 5, 2018
A couple of points, in addition to the above:
#1. I prefer qualitative evidence to quantitative. In other words, I'd want to understand how a money manager's trading strategy works: what his 'edge' is, (especially) how he manages risk, and satisfy myself that he's fully conversant with the process of trading; in addition to seeing a performance history. There is always the possibility that TEs or myfxbooks can be faked or manipulated, or the results propped up by unsustainable MM (e.g. martingale variants; adding to losses) that will eventually fail, or some other ruse.
#2. I would NEVER allow a trader to manage my money unless:
He met the requirements of everything stated in point #1;
He provided evidence that he was both fully accredited, and regulated, by the relevant governing authorities in his country;
I knew the physical address of his operation;
I'd seen some kind of third-party verified performance history showing consistent results over several years' worth of changing markets;
The funds were held in a segregated account, in my name, and he merely had limited POA allowing him to trade on my behalf.
Risk model calculator.xls
Glad you started this.
I would add that not only is large drawdown a sign of problems but very large outsize profits are also a sign of possible trouble ahead. If you are gaining forty percent in a week you can lose just as quickly.
Excellent work David.
I think you are right, it’s important at the outset to determine one’s “performance agenda”. Similar to you, mine is low risk, high return and very importantly longevity.
Once this established, it’s a matter of matching this “performance agenda” with the appropriate metrics for analysis.
At FF and at most forums, the agenda seems more closely aligned to high risk and high return; longevity is ignored, or what is required to secure it, is misunderstood.
A recent exchange with a member illustrates this point. He had found himself a “trader” who in one month lost 50% of his $100k account, and in the next made 100% on his account. The member was unable to see that the DD spelt high risk, and equally so, the profit. The concepts of risk of ruin and longevity were lost on this member. To cap it off, our star trader was using a demo account, a variable which the member didn’t acknowledge as significant; which opens another can of worms.
Nevertheless, unless one is able to articulate their “performance agenda” and then match that with the appropriate metrics for analysis, it will be difficult to understand why a TE or other account, whilst looking impressive, is actually headed for disaster.
Thank you very much Hannover. Very good explanation and very interesting.
Just reading it, i only want to observe about
2. Profit factor: since you say "IMO a better PF is obtained by dividing the profit pips by the loss pips. That cuts MM (point (b)) out of the equation", it is true but only to a certain point (tell me if i am wrong)..If let say:
- trader A enter a trade with 4 lots , and exit the trade with 25 pips, it will be credited with only 25 profits pips
- trader B enter same trade with 4 times 1 lot, and exit the trade with 25 pips, it will be credited with 100 profits pips...so in that case MM is the same, but trader B looks much better in terms of PF...
ps. i am missing point 1,2,3 (i guess) and 4 of your list...so i am not waiting for your fund transfer
@PeterE: Yes, I saw your exchange with mmaker in TPOTrader's thread, and visited the "star trader's" website to check his performance. He lost around 25% on a single trade in December: he'd set a 70 pip SL only to see price (EURJPY if I recall) gap over his SL, during the weekend, to create a loss of 195 pips, LOL. I didn't realize it was a demo account until somebody pointed it out in the thread (trading those vast position sizes with apparent nonchalance did make me suspicious, however).
@jeronimo: Your assumption is correct if the trader makes only positive pips (winning trades). But if Trader B (in your example) is making 4x the number of pips on his winners, then he will also be losing 4x the number of pips on his losses. Hence the ratio of winning pips to losing pips should sort out any anomalies. That's one reason why I prefer ratios (A divided by B) as opposed to differences (A minus B).
However, as a general principle, you're correct (IMO) that the summary statistics can be skewed in a way that makes the trader's performance appear better or worse, than what he actually is. That's why I believe that it's also important to study the individual trades. It's surprising how much you can learn from these; sometimes it's possible to even decipher parts of his trading strategy.
I show here an example for classroom case study. Unfortunately I don't have it in TradeExplorer form, but the details are available:
I agree that PF>2 is a good threshold to look at.
If someone can achieve PF2 consistently for a long time, I will see he is extremely successful.
Of course, some will show accounts PF>100, but believe me these accounts are fake or by luck or with extreme small sample sizes.
Notice that people like to show their performance when they are extremely good, and like to hide if they are bad or not extremely good.
This kind of survival bias is popular and can fool newbies for a long time.
Nice and very informative thread. Thanks for sharing your views David.
A nice TE with 10 or 20% profit, then next time I check it out, it's gone.
TE back in profit and it re-appears, then later, gone again.
It suggests to me that these people consider it more important to appear to be a good trader than to actually be one.
A couple of glaring examples:
'A good trader should never risk more than 2% of his account.'...........Says who?.....and why? What criteria was used to arrive at this grand sweeping statement? For instance....a trader with a win/loss rate of 55/45 would more than likely be uncomfortable with a 2% risk. A trader with a win/loss rate of 75/25 would perhaps be comfortable with a 2-5% risk.....maybe more depending on the situation. Does the criteria take into consideration the type of trader who never lets his account exceed (say) $10k before making a $5k withdrawal? After doing this twice, he is trading purely from the winnings and he must (obviously) have a good system and win/loss rate so he may choose to risk 10% of the account.
There are many, many permutations that we could consider.......or shall everyone just keep parroting the same old tired cliche out of ignorance and laziness?
What about: '98% of people lose by trading FX'.......Again....says who? What does 'lose' mean? Lose the account from the get go? Lose the account after making back 100% of the investment?.....Lose a trade?.....Quit FX trading because they find it's not their cup of tea?.....Lose the account, or even more than one account before becoming incredible skilled and successful? You see?.....once again, there are many permutations that need to be looked at before one could even dream of considering a set of criteria that is meaningful. Or....again.....it's just easier to keep parroting that tired old cliche out of pure ignorance.
I honestly appreciate what you're trying to do here Hanover, honestly I do. But there simply isn't a 'one size fits all' glove in FX trading, there is (and I know you know this) much, much more to it than that.
Just as people scoff at TE being only displayed when they are in profit, I scoff at the repeated cliche's that are just parroted without any validating research to provide some substance to the (repeated) claims. When I see this kind of nonsense repeated as if it was gospel, and the poster cannot show any validation of the statement including the criteria used......then I very quickly lose interest.
Profit Factor is important also in a creative process of building "automatic" trading system. (It is not "automatic" in regular understanding, instead it is very sophisticated piece of programming and mathematical art).
The problem : the algo for your trading system is not described in a book. This is unusual situation for any regular programmer. There is no theoretical proof for your algorithm.
Therefore, if you write a program for automatic trading the Profit Factor obtained from backtesting is actually the first and most important value. The drawdown is the second.
The next problem: PF obtained from backtesting is subject to a row of coincidences between the splashes on the market and (optimized) parameters of your automatic system. Therefore backtested PF usually 2-4 times HIGHER than obtained from real world trading.
from what i observed trade explorer confirms the 5% rule. most explorers show failure and the small comeback and they enter the top but that is not ok. the ones that do go up for a long time have huge problems. i haven't seen them all yet but i will in the future. an observation for now: if i see thousands of trades and a chart constantly going up in a straight line that would be a sign to get out. i've seen at least 10 examples of a straight line followed by a quick drop and eventually a failure. a healthy chart should always find retracements . key aspect is how much they are risking aka average loosing trade relative to what winrate and risk reward allows. profit factor is not computed correctly in the trade explorer. profit factor is average win* winrate/ (average loss *(1-winrate)). i found very few correct profit factors calculated. you have to do this yourself.
let's look at this example. http://www.forexfactory.com/sir*oliver#49-report . the profit factor is not 6 should be 1.15 and the chart is a straight line up . after just looking at the chart this should go down in the future but there are some factors that suggest it's different from the other 10 i've seen.
another similar explorer most likely to fail http://www.forexfactory.com/mechie#77-graph
An excellent post, I rad & admire all your writings as well as your coding, I used the Daily Line Boxes many times.
I know that you are terribly busy trading & coding, but I was wondering if you could take a look at the Trade Explorer linked below and give me your opinion where this might fail in the future?
I know it fails your very first Point #1 of 1 years data as it is only 2 weeks going now.
And Point #2 - Profit Factor less than 2 but probably was because I started out with equal R:R as TP / SL the same (1 ADR) - I have now tweaked to 1.1 * ADR for TP, 0.9 * ADR for SL to see if that can push the winnings up a notch.
Which I guess affected Point #5 - Average Win to Loss ratio
Point #3 - Draw down, I think it is OK there.
Point #4 - I am using a position size, based on the stop loss, of 0.5% per trade, maximum of 2 open trades per pair opened at least 4 hours apart.
So, if all 28 pairs had 2 open trades, and the complete FX market reversed all at once, potential could lose 28% in one big hit; keep in mind, my main reason for testing the EA in this fashion was to TEST the functioning of the EA, so far so good.
It has managed to run up a good profit buffer thus far, so this may have legs do you think?
Much appreciated, E.
So, according to point 1, the most important for you, a trade explorer with 12 months and 500+ trades is statistically valid, even for a martingale ?
I ask, because this is my case, I off course would like to achieve those 500+ trades but in fact my goal is to forward test live my martingale for 5 (yes, five) years ! This should normally give my 3000+ trades.
It's just what I have started to do, if you are interested in following that experiment...
Sorry, I have been intermittently absent from the forum, have just noticed these 2 month old posts now.
@Erebus: I wasn't able to find the link in your post.
@broketrader: Martingale doesn't fit with my initial criteria ("steady long term account growth with tight control on risk"). The first criterion is met, and most comprehensively so, but at the expense of the second. In my world, Martingale represents a different kind of approach to trading. It may represent a good way to try to grow a small account — the loss of which can easily be shrugged off, if need be — into a large one.
I don't believe that it's possible to accurately test for extremely rare events ("black swans"). The margin of statistical error is simply too great. To explain with a simple example, how many years would you need to drive without having a car accident, before you can say with confidence "I will never have a car accident in the future"? IMO martingale fits that type of category. Given enough time, every price pattern that can possibly occur, will eventually do so. The probability of the 'death sequence' occurring may be tiny, but the real question is what is the extent of the damage that will ensue, if it does occur during the trader's career. Then it comes down to a personal decision: what kind of money is involved, and is the trader ready to take that kind of risk?
The most intelligent use I've seen for martingale is a breakout strategy: price always eventually breaks out of prolonged congestion, the only question being how many false breaks occur before the profitable one, in a worst case scenario. I was going to code an EA for this trader (at no charge), until I discovered that the job was going to take several weeks. It is a profitable strategy, and like all martingales provides a steady income, but again it comes down to whether the individual trader is willing to embrace the risks involved. I wasn't, and hence I backed out of the project.
I have been trading live for 2 years now. The first year i write off as it was a disaster but the second has proven to be good. I had to withdraw money as i needed it but now starting to build it back up with salary topping up monthly. I currently have a profit factor of 12.73 and an 80% win rate. I have linked my account to my profile so you can have a look at the details. I am serious about taking this full time at some point and will be injecting money every month to build the account side-by-side with trading profits. I am reaching out to full-time traders or anyone working in the commercial space for advice as to what to do from here in my planning to take this full-time. In all honesty i never really expected to get this far but i have given it solid focus to see where i could take this. I initially thought if i could make better returns than a savings account that it would become a pension building tool but think i might have what it takes to make it full-time.
My approach to trading is 100% fundamental and happily enter and manage trades without looking at a single chart. I believe this is a sustainable way to doing this with no risks of the strategy not working at any point like you find with a technical system.
© Forex Factory