Jul 172012

Many traders on Twitter are using stats for their trading.  One of the simplest examples is the gap fill stat.  Many years ago traders would fade gaps because they were often filled.  That’s changed in the past few years so be careful if you do that.

Lawrence Chan has been a pioneer in this area and he has written a couple of interesting books on using stats in your trading, specifically two are for the S&P 500.  Before you rush to purchase these, I suggest reading ahead.  I think the books are worth their price, but they’re not the holy grail.  I believe most of the stats are not very applicable and I’ve found one to be invalid.  More on that later.

At first I was intrigued by the stats in the first S&P book (the 2nd is relatively new).  Some of them have 75% probabilities.  So easy right?  Just put on the trade and you have a 75% win rate.

Not so fast.  Trade expectancy is based on two levers which are the Risk-Reward and the win rate.  Expectancy of a trade is it’s reward * win rate – risk * loss rate.  So if we have a target twice as big as the stop, the R:R is 2:1 (we usually right the reward first even though we say “risk-reward”).  If the win rate is 40% then the expectancy is:

2*0.40 – 1*0.60 =  0.20 pts/trade.

So what can a trader control?  We can only control the risk and reward.  We can’t control the win rate.  However, the market is efficient in that if we take random trades, the win rate will be exactly the win rate required for us to end up exactly zero*.  I’ve proven this with random trade simulations and it’s quite fascinating.  And logical if you think about it.

So how do we make money?  We make money by controlling the R:R and getting a win rate above random win rate.  If we use 2:1 the random win rate is 33%.  So to be profitable, we must be this.  In our example we had a positive expectancy of 0.20 pts because our win rate was slightly higher than average (40% vs. 33%).  The edge was effectively slightly less than 1 tick, which coincidentally is the edge that I’ve often observed in my own trading (I wrote quite a lot about that over the past couple years).  Don Miller, a famous retail trader, has a similar edge.  It’s hard to get a win rate greater than random in something as unpredictable as the financial markets.  And I think many traders may only have an edge of a few ticks.  That means to make any reasonable money, one must trade size.  Which is exactly how Don Miller makes millions.  But trading size is not as easy as changing the order quantity on your DOM, but that’ll be for another post.

So back to our stats, this is is why having inverted R:R where risk is bigger than the target is ill-advised.  If our R:R is 1:2 we need 66% win rate.  And that’s difficult to do.  If you trade with very large stops or no stops at all, the required win rate is 85% or higher.  And that one loss will wipe out all the winnings.  This is why moving stops and cutting winners short does not work.

So how do we get a win rate above average?  We don’t enter random trades.  We enter only when we have a well-defined edge.  I say “well-defined” because intuition doesn’t work very well with trading.  When we think we have an edge, we’re often being suckered by someone who knows they have an edge.

So to be well-defined we have to be in a situation we’ve seen before, something repeatable.  And that’s where the trading plan comes in.  This has been a constant challenge for me because my trading was always so complex and had too much discretion, and therefore was not repeatable.  I’d have a good month followed by a bad month.

On the other hand, a simple setup or pattern that anyone could do isn’t going to work.  The market is too efficient for that.  And so coming up with your well-defined edge is a long trial & error process.

Stats can be used to help increase the win rate, and I’ll write more about this in a future post.


*I didn’t include the impact of commissions which will gaurantee any random trading will result in a loss.



  12 Responses to “Stats, expectancy, and the levers”

  1. …” this is is why having inverted R:R where risk is bigger than the target is ill-advised.”
    I don’t agree with this statement and your D. Miller example contradicts it, too. He has often said that the statement above belongs to the trading myths of the industry. There are setups with a high win probability, but a R:R which is inverted (Miller for example has an inverted R:R as he has written).
    In the end it is about finding a way of trading in which you really have gained confidence and then you can accept the losing months which are not avoidable and you can increase size too. Easy said, but hard to realized.

  2. Markus – do you know of any other traders who advise having an inverted R:R?

  3. I think most scalpers or short term traders have it. For swing traders it might be not possible. And classical momentum is just the opposite way: high R:R and low probability.

  4. A long time ago (11/2010) FT71 had some things to say about scaling on twitter. He specifically talks about trading a 1-2 lot. I saved it to a notepad file for reference (it became engraved in my mind). I also have a bookmark from that time period for a url that has all the scraped #FT71 conversations (I have no idea who is behind it. When FT71 was active on the stream I would plot his trades and tweets on a chart so it was a good tool). Anyway I think you will find it VERY interesting as you are involved in the Q & A.

    Reads from bottom to top:

    As I mentioned in a previous comment a couple days ago 3 NQ > 2 ES. That extra scale makes a HUGE difference on a number of levels. You’ll be able to hit singles and doubles more consistently.

    One last thing I would like to add: you said “intuition doesn’t work very well with trading”. I semi-disagree. Superior intuition in the form of context is the discretionary traders x factor. I was a winning poker player at decent stakes until the feds blacked out the US. The game has an incredible amount of layers to it. When I first started I SUCKED! Then I started to learn about positioning, hand ranges, theoretical equity, exploiting player styles, tilt, etc.

    Did you know going all in with 7,2 (worst hand possible) makes just as much sense as going all in with AA at certain points in tournaments? I got to a point where I was playing 15 tables at a time and making the majority of decisions in under 1-2 seconds. Intuition and loose math/strategy started to blur together. I got to that level by studying my losing hands (not the math). First I would look at the context. Was I tilted? Was my opponent tight, aggressive, loose, or passive? Is he a winning or losing player? Past hand interactions? How many times does he raise from X position? How many chips does he have? How loose are the players that act after me?

    Those are the kinds of things that go on in FT and Robs head. Knowing that helped me a lot. I know you used to play poker for fun so maybe that will help you as well. Poker is efficient just like the market. Hands are distributed equally. Every player is dealt the same cards.

    I’ve babble a lot but I hope my point came across. Cheers.

  5. Thanks for sharing S.
    “2010-11-26T10:14:59|@cunparis|@FuturesTrader71 So in my case I use 1.5 stop so I scale 1.5 & then go for 2.5? WHen consistent I could add a third for the gravy. #FT71”
    If you take this example you have a max reward of 2 points if you trade 2 lots. If your risk is 2 points aka 8 ticks you have a R:R of 1. But your average Reward will be less than 2 points cause sometimes you get only 1.5 points and dont get the fill for 2.5 points. This makes a Reward:Risk of less than 1.

  6. In term of stats, Kevin at StockFetcher forum has a filter for gap up and down with a very good backtest result. So far I do manual back testing and it is almost 100% correct. But the execution is very hard since it need to filter 2000 stocks and order 5-10 of them at the open as fast as possile.
    I am ordering the book Optimize your trading edge of Bo Yoder cause one of the best member of Eminiplayer uses it at the R:R guidelines, I hope I can learn something from that.
    Quang (one of your first customer, I hope you will be back with it soon)

  7. Scaling out of trades will lower your R:R as Markus said. That is why it needs traders like FT71 with a high win rate to be profitable with scaling out.
    What is the difference between a profitable and a loosing trader? As you said Michael, it’s the win rate, but compared to other traders with the same method resp. timeframe.

  8. I think everyone has to find what works for them. if there are scalpers with inverted R:R making money then that’s great. I think the vast majority of traders do not recommend this and I haven’t seen any scalpers making money this way, and possibly don miller is an exception. Even if don miller is an exception, he doesn’t teach his method unless you pay thousands of dollars and looking at his site I haven’t been able to figure out anything he’s doing. so without Don Miller I don’t see many people using large stops and tight targets. But I’m not going to say it’s impossible.

    FT can take small scales but he also has small risk. very small. sometimes a few ticks. so he can take 5-6 ticks and will have a “positive” R:R.

    I once ran a stock system based on stockfetcher backtests and my real money results going forward didn’t match stockfetchers fills. it is very generous on fills, especially if you’re taking the top x stocks from the filter and does things that one can’t do with real money. SF knows which stocks will fill but in real time you don’t.

    All interesting discussion but the main point I wanted to make is we control the R:R and the market controls the win rate. and I’ll talk about stats soon..

  9. An interesting article anyway. Looking forward to the next one.

  10. I admittedly did a very bad job at explaining what I was thinking.

    Hypothetical scales = multidimensional risk.

    Lets say you trade a 2 lot with a 1.5 stop. Your explicate risk is $150.

    IF you were to scale 1 lot every single trade at 1.5pts the explicate risk set at $150 is no longer accurate.

    If you can get that first 1.5pt scale just 50% of the time your actual risk per trade is $112.5 not $150.
    The 2:1 , $300:$150 (12 total ticks) shrinks to 2:1 , $225:$112.50 (9 total ticks). Obviously it’s much easier to capture 9 rather than 12. Out at 1.5 and 3 gives you a 2:1. Plus you have the 1.5pt cushion.

    2 Contracts
    $25 per tick * 6 tick stop = $150 stop out
    1st scale = $75 (6 ticks).
    ($150 * .5) + ($75 * .5) = $112.5 implicate risk.

    ($225 * .4) – ($112.50 * .6) = $22.5 +EV (slightly less than a tick). However, the expected value will also be multidimensional. If you hit your 1st scale 50% of the time how often do you scratch if it comes back to entry? How often do you scratch at theoretical average? How often do you add? etc. It could very well be that your EV is actually 2-4 ticks (imagine this with a 50% or 60%+ hit rate). This grows in complexity at 3+ contracts (in your favor).

    1.50pt scale 40% of the time, risk per trade is $120 not $150.
    1.25pt scale 50% of the time, risk per trade is $118.75 not $150.
    1.25pt scale 40% of the time, risk per trade is $125 not $150.
    *Of course as you shrink the 1st scale it adds to the ticks you need to capture with the last. So 1.25-2pts 1st scale is the ideal balance. Also because the average harmonic rotation is 2-3 pts you put yourself above a pull back.

    To grind out consistent profits and water down variance it makes the most sense to scale early (unless you can emotionally handle NOT locking in profits and the variance that goes a long with that) even with a 2 lot.

    I really wanted to clear that up. Thanks Michael.

  11. You have to consider that the 1st scale trade must have a very high win rate. Good traders seem to almost always reach the 1st scale. In this case scaling out is good and will lower your risk. The 1st scale is nothing than a scalp, and scalping without a high win rate is difficult.

    As soon as your win rate on the first scale is lower, scaling becomes a worse or loosing strategy.

  12. I agree Berti. Like Michael says in the above post “However, the market is efficient in that if we take random trades, the win rate will be exactly the win rate required for us to end up exactly zero*”. Thus in theory the 1st scale should be hit at least 50% of the time. Even at 40% the strategy will still work.The last contract can be exited at 2.5-3.25ish (decided on profile/setup & 1st scale hit rate) points for a decent 2:1ish trade.

 Leave a Reply

You may use these HTML tags and attributes: <a href="" title=""> <abbr title=""> <acronym title=""> <b> <blockquote cite=""> <cite> <code> <del datetime=""> <em> <i> <q cite=""> <s> <strike> <strong>