Curious, has anyone tested different settings on the lookback periods... IE, what happens if ratio lookback 14, stretch 14 are set to 20. Wont extending the lookback smooth out the ratio moving average and produce less trades. Nick
365 days is kinda short period, also need to make sure you have enough sample size (at least 30 trades) to make sure you findings are statistically significant. I have a pairs trading algo coded in WealthLab so I can run some testing for you, PM if needed. Just wanted to bump my question regarding intra-day pairs trading, anyone currently doing or thinking of doing so?
FYI a couple of trades Friday and Today (Monday). I won't clutter Jonny's journal with my trades, but someone thought my use of expected ratio and expected profit was interesting. My trade size is $2500 each side. A pair entered Friday: Long CS @ 30 Short LYG @ 4.29 Ratio 7 Expected ratio 9.51 for a profit of 9% Closed CS today (without a signal from PTF) at 27.83 for a loss of $185. Letting LYG run with trailing stop loss, currently a profit of $400. A pair entered today: Short CCO @ 3.31 Long HHS @5.38. Ratio: 0.615. Expected ratio 0.560 for profit of 4.5% Walt B
HI Walt and Johnny, thanks for the continual great info. Just wanted to say Walt that I too am having some problem with PF regarding false signals, negative correl and a few other things. I've made Jared aware of them so hopefully he is working away at them. I've been trading with PF recently but it seems to have all fallen in a heap, so I have to analyse what I'm doing wrong. One thing I was trying to assess was cutting the losing leg and letting the winning leg run a bit which I think Walt is what you were doing in your last trade. But, I'm not able to easily backtest this. I'll keep plugging away, but I must admit my confidence has taken a bit of a hit ! regards Ivan
Hey Ivan, I received a quick response from Jared at PTF. I followed his instructions and my data base is rebuilt and working MUCH better. The problems are corrected. Here is what he told me to do: Walt, Excellent to hear you are profitable! Keep up the great work. With the ratio charts it doesn't always show every day, because it has to compress the data, however it shouldn't skew as much as you describe, to fix this, close down PT Finder then open SQL Server Management Studio Express(will have to download if you don't have it), it can be found in start menu -> programs -> SQL Then click on new query and copy and paste the following; USE pairTradeFinder; DELETE FROM tradeData; DELETE FROM stockTradeHistory; UPDATE stock SET lastDownload=null; Click on execute. Then open pt finder again. Basically this deletes your current trade data and reloads it, all your stock groups and pairs will remain intact. Kind Regards, Jared Mann General Manager www.PairTradeFinder.com
I'm a big looser, without the hedging side in place. At least I've been in the past. I try to remember that the reason I'm pair trading is that without both halves of the pair, I'm a looser. So it's important to let both halves run and if you want to trade pairs by the book, then both need to run until their ratio reverts to the mean average of the ratio. But when one side is really profitable (for me with a $2500 investment each side, that's over $250), I'll consider cutting my losses, rather than being too greedy and hanging on hoping the loosing side will come around. The $250 profit gives me enough space to set a stop loss and come out to the good. I hope. Also, as the trade is nearing it's end, and the ratio is approaching mean average, I'll get out of the loosing side, and set a stop loss on the winning side. I don't see the wisdom of riding the looser to the end of the trade, but I don't give up the ship too early, either. I'm no expert, but that has given me some additional profit and I don't think I've missed many opportunities. I CERTAINLY don't want to close out my winning side just because the software gives me a signal. That signal is only to tell me where the ratio is. I'll take continuing profits on the winning side just as long as I can. Ivan, I don't see what's to back test on this. When you have your profit on one side, instead of closing it on the signal, simply hang on to it with a stop loss in place. That stop loss needs to be set properly, but you don't have much to loose, and maybe a lot to gain. Friday, after taking some losses and having data base problems, I lost some confidence, and was feeling down. Today the trades I entered last week are profitable (about $800 worth), I have stop losses in place, I picked one good trade today, and I'm feeling "in the money" again. When you fall off the horse, you need to hop back on. I think. WB
Thanks Walt ! I'll download it and give it a try, it might clean up some of my other problems I'm having. I've always wondered if there was some dbase "housekeeping" that was required as the thing just seems to get sluggish over time. I've actually split my pairs over 2 PC's to try and keep it lively regards Ivan
Failed at it last summer. I was doing a convergence on famous pairs: CVX:XOM, T:VX, ... FNM:FRE (back then, lol). I had settings like RSI, Force and MACD histogram. Most of the time The where just keeping going on their way or I was getting out too soon and the net P/L was negative. May be back at it with different settings/ideas and try to insert a divergence component. So I dont really have any insight :eek:
Couldn't agree more yobo, keeping it simple is my motto, and yes the ratio charts are what I mostly look at too, however I find it interesting to look at other filters out of curiosity and see what could be enhanced. Its all about continual improvement and consistent application.
You can either factor it into your trading model or simply not trade stocks with a dividend due in the next 2 weeks, go to www.earnings.com to check for upcoming ex-dividend dates. I prefer not to trade around dividend time unless I can go long the dividend paying stock and collect the dividend, also there is a proven tendency for stocks to outperform the market by 2-4% in the 4 weeks preceding the ex-dividend date, the probability and payoff of the event rises with the amount of the dividend and the franking.