Pair Trading Strategy Journal

Discussion in 'Journals' started by jonnysharp, Aug 18, 2008.

  1. thanks. Keep talking about the RSI. My paritrader is broken right now so i've been exclusively using a system i built. do you know what the formula is for the RSI calculation? i tried creating one that is simply trying to plot a 14 day RSI of the spread...but i'm not getting a chart that looks like yours for the RSI.

    Anyway, today i closed out two trades at profits.

    Nice profit on SYT/MOS.
    Tiny profit (just above breakeven) on IWV/VTI but i would have expected that based on the chart (little spread).

    I have quite a few badly underwater so despite my 7 profitable closes to 0 losses so far once i start closing these losers out if they don't turn around i'm afraid my profit won't look great; but we'll see.

    wasn't back in time to scan for new ones before the close but a few i see after hours that i may get into tomorrow morning are:
    VALE/TS
    CL/PG (correlations are not quite great though)
    WMB/STR
     
    #1781     Jan 14, 2010
  2. waltbx

    waltbx

    Here's one that has earned me 2% since yesterday and has a long way to go yet. My estimate was 3.8% profit at 50% reversion to mean to 7.6% at 100% reversion. It's been on my radar for awhile and I waited until RSI reversed.

    RSI peaked at 93 (well above my 70 requirement) on the 13th and is now reversing. Good sign.

    The ratio is on the good side of the trend, and is reverting to the mean.

    Short GLT / Long TIN. One of the best pair charts I've seen in awhile.

    (Past performance is no guarantee...)

    Walt B
     
    #1782     Jan 15, 2010
  3. waltbx

    waltbx

    Hi Mike,
    Sorry, I don't have formulas for RSI, but I've seen them on the net. I use the charts in PairTrade Finder exclusively. Quick and simple. But the RSI values seem to be off a bit from what some of the guys calculate in their spreadsheets. I don't think it matters much.

    I looked at the three trades above and the only one that I think is ready, from the RSI charts, is CL/PG. RSI is 24, Ratio 1.31, Mean 1.35. I'd wait until the RSI reverses.

    Yet, this pair will tie your money up with only a small potential for profit. I have a rule of thumb I use to quickly guess on the profit potential. I want the difference between the RATIO and the RATIO MEAN AVERAGE (I look back 17 days) to be 6-10%. In other words, if the Mean is 1.35 (as it is with CL/PG), I'd want the RATIO to be 0.081 - 0.135 away. This pair is only 0.04 separated, too small of a potential to tie up my money, and for the risk. In your case, when you are taking a larger number of smaller position, this may be ok for you. But in the end, it is still a smaller percentage spread over your total account.

    To make it easier for me to make a quick calculation, I look at the ratio and simply move the decimal one place: CL/PG has a RATIO of 1.31. I want a separation of 0.13 more or less, to the MEAN. It does not meet my criteria.

    I've picked these numbers to give me a potential profit of 1-2% if the RATIO reverts at least 50% back to the mean. In the last few months, my expectations have lowered and I'm sometimes trading with ratio/mean separations of 5%, instead of 10%. Last spring, my expectations and results were three times higher.

    The other two pairs have RSIs that aren't close to 30 or 70 nor have they been recently. So I wouldn't risk my money on those.

    That's my thoughts. It'll be interesting to see how these pairs pan out.

    Walt B
     
    #1783     Jan 15, 2010
  4. Walt, thanks for the info. i think i am slowly getting comfortable picking the right ones again as I think i did a better job early last year. i think using RSI or some other ideas i have to make sure there is enough profit makes sense.

    For what it's worth I closed out 4 trades today. 3 due to an exit signal, one that wasn't quite there yet but the profit was nice so I decided to jump.

    Clsoed trades:
    TR/ITW: small profit
    LFC/CEO: one i closed out a little early; nice profit
    VDE/DIG: small loss
    XES/IEZ: small loss
    USO/OIL: scrath; counted as loss

    So far: 12 closed; 9 wins (75%), 3 losses (25%).
    I was around 70% last time so this is pretty consistant and I will have more losers soon.

    So far the value of winners to losers is heavily stacked to the winners but that will change soon too.

    I am probably going to add some screening to my entries soon but went with a new group today that looked good at the end of the day:

    V/MA
    VALE/TS (most questionalbe of the batch to me)
    TSL/VLO
    SNH/NHP
    CRZO/NFX
    NNN/OFC

    Mike
     
    #1784     Jan 15, 2010
  5. waltbx

    waltbx

    Mike, it would be interesting in seeing a .pdf of your trading results over that last month or so.

    Walt B
     
    #1785     Jan 15, 2010
  6. Walt -

    I'll share once i have more data; maybe in a few weeks Thanks for your ideas on profitability.

    In thinking about your approach to narrowing down the list and being more selective it occurs to me that my parameters have been:
    * high correlation over time
    * fairly horizontal channel with ups and downs
    * trigger (spread average outside of 2 Standard Deviations)

    I think that making sure the spread is there makes sense. Ignoring that is probably why I had a high quantity of wins but low value in wins; high probability trades but low expected returns.

    i'm going to work on some updated charts to help me track what you are tracking. i also managed to fix my pairtradefinder software so at least i an use that too.
     
    #1786     Jan 15, 2010
  7. walt,

    2 things.

    First, i went back and created some new indicators for myself incorporating the RSI (which is a bit different then pairtradefinders for some reason) and profit potential.

    One thing that is interesting. I went back and looked at a spreadsheet where i had tracked all of Johnny's original trades and I eyeballed about the best 15 profit trades he had...and all of them had profit potential of at least around 5% and RSI of > 60 or < 40.

    So I am beginning to agree with you that looking at those factors is very useful and will start thinking about those next week (i actually added them as items in my scan so i can see them quickly).

    two questions about your post.

    when you talk about a ratio mean average w/ 17 days lookback are you saying a moving 17 day average of the ratio? just want to make sure i'm understanding the terminology.

    also, any reason you pick 17 instead of, say, 14 or 20? or just splitting the difference?
     
    #1787     Jan 15, 2010
  8. Update: Yesterday I closed out kb/shg (from 1/6) for a modest profit.

    did not have time to scan for new trades.

    So far wins 10, losses 3, and wins about 3x size of losses. That will change quickly when some of my losers start to close out.

    i have created a few new charts and incorporated into my scans Walts thought on profit range (big enough to make a difference) and RSI (at one extreme or the other).

    I'm not sure yet if i'll let them influance my entries but i may. I have also gone back to my earlier trades and in my spreadsheet am tracking those that would have met the more stringent criteria vs those that would not have to see which do better or worse collectively.
     
    #1788     Jan 20, 2010
  9. Closed out some more and entered more today.

    CLOSES:
    NSC/CSX; modest profit
    KB/SHG; modest profit
    cpt/hst; small profit
    str/hal: large profit; best yet
    hiw/dlr: modest loss; largest loss so far

    SO FAR:

    17 trades; 13 wins (76%).
    Average win size for $1000 trade: $24
    Average loss size: $12 (this will be going up for sure)
    Average days in trades so far: 8


    If you assume no margin and that when I buy/sell $1000 I use the full $1000 in capital (i actually probably get more leverage due to cash from the short sale so this is worst case scenario)... I'd calculate the return so far as follows:

    250 trading days / 8 days per trade = 31 round trips
    31 round trips x (76% win) x $24 = $565
    31 round trips x (24% losses) x $12 = $89
    Net profit: $476 = 47% with no compounding

    Don't know if anyone has a better way to calculate return.

    I know my average days per trade will go up slightly and my average loss size will go up significantly.

    I'm thinking if my win rate drops to 70% but my average loss goes up to $30 (from $12)...then I'd be at about 24% as follows:

    31 x .70 x $24 = $520
    31 x .30 x $30 = $279
    Profit = $241

    Meanwhile the real return based on those #'s would be higher since the entire $1000 is not allocated.

    Also, if I begin to incorporate some of Walt's thinking and look for those w/ a higher profit % my average win should go up.

    On oddity, although i'm still early. When i did this last year my average win was $51 and my average loss was $86. Both #'s are much higher than now so I'm not sure if i'm selecting stocks differently. At that time I had 69% wins.

    Finally, entered into 4 new ones today:
    UN/UL (lower profit target)
    APA/WMB
    PPS/CPT
    EMR/PH
     
    #1789     Jan 20, 2010
  10. saico

    saico

    The problem that I noticed is, that 1-2 big hits between 100 and 150 USD that will come very likely during a month reduce the profit dramatically. Sometimes to 0, or below. Even with these positive stats in your back its often a zero sum game.
     
    #1790     Jan 20, 2010