No, the point is that adjusted prices do not exist in the real world, adjusted prices are a convenience for the trader that lets him account for dividends and splits. So it doesn't matter that the price data wasn't being.adjusted in real time. It's a BACKTEST! As long as there's no future leakage, then it's no harm, no foul. Once more, if the adjusted data is good enough for calculating the buy-and-hold result, then it's good enough for generating timing signals in any other trading strategy. Buy-and-hold is a trading strategy afterall. It really doesn't make sense to believe that adjusted data is good for one and only one trading strategy. Try wrapping your intelligence around that basic concept.
Close vs. Adjusted Close
- Thread starter craigatelite
- Start date
I suggest you try some backtests with dividend adjusted data and without adjustemnts. The rest is senseless talk.
It seems you have not done any comparison work. Buy and hold is not even a backtest in the full sense of the word. It is just subtracting two prices.
Aggression will get you nowhere. I am trying to help you and you become aggressive. Apparently you have done many mistakes backtesting and you are trying to justify them by being aggressive, just like a little misbehaving kid.
You are now on ignore because it is a waste of time dealing with you.
Back to school kut2k2. Let's do some arithmetic. There is this new stock IPO and my system will buy if Price > MA(3).
The stock starts trading and it registers the following price series in the next 3 days:
4, 5, 4
obviously, MA(3) = 4.3333 > 4 and there is no buy signal at the close of the third day.
Now, just for illustration purposes, let us assume that this unique stocks pays a dividend on day 4 equal to 2 (what a wonderful world that would be!).
According to the way prices are normally adusted by various vendors, the 3-day series will become:
2,3,2
Still, MA(3) = 2.3333 > 2 as expected and there is no signal.
Now, on the forth day the stock trades at 4.
The adjusted series is 2,3,2,4
MA(3) = 3 < 4 and we get a signal to buy the stock based on adjusted data series.
However, according to the unadjusted series:
4,5,4,4
MA(3) = 4.3333 > 4 and there is no signal.
In other words, the dividend adjustment created an artificial gap in prices that boosted current price in relation to the MA, generating a false signal.
Do you like false signals kut2k2?
If you don't, I suggets you do some refreshing of your arithmetic skills.
And once again some goober proposes contrived nonsense as an excuse to reject reason. Why am I not surprised?
I for one would never trade raw Price over any MA. Ever heard of whipsaws? My system needs at least 24 data points before it takes off, so there's no chance of going off halfcocked with insufficient data to generate a decent signal. YMMV
Use only real prices when you backtest if you want to be in touch with reality. Never employ any data adjustements. Adjust entry price and # ofshares in case of a stock split and roll contracts in case of futures rollovers. Use only programs that allow you to backtest this way. Trading Blox comes to mind.
If you hear the words adjusted data run, run, run. Idiot alert.
If you hear the words adjusted data run, run, run. Idiot alert.
How about recalculation all prices so all prices are adjusted and using this adjusted prices for back testing? If I use original prices and there is split somewhere, there occurs sudden price jump which destroys technical indicators etc.
So I do multiple all prices (open, high, low, ) by ad_close/close
This prevent unnatural price jumps in data. Im I correct?
So I do multiple all prices (open, high, low, ) by ad_close/close
This prevent unnatural price jumps in data. Im I correct?
+1
It makes sense to me. It is also much more realistic in terms of commission cost. IMO intraday traders can use unadjusted data anyway with no problems since any adjustements are done at the open or close of a day. Position traders should track splits or rollovers and adjust any open positions. That is not very hard to do. Yes, I agree, real traders should use only real data. Nothing else.
Clearly adjusted prices ares not appropriate in backtesting for the reasons mentioned above. This said, what do you do when your system compares any 6-12 months trend indicator (ROC, moving averages, etc) across several different stocks?
Conceptually you need to include dividends in the calculation of total
return. It would be unfair to exclude them penalizing stocks which had
larger distributions. However, I also get the point made by the article that dividend adjusted data distort backtesting...
Conceptually you need to include dividends in the calculation of total
return. It would be unfair to exclude them penalizing stocks which had
larger distributions. However, I also get the point made by the article that dividend adjusted data distort backtesting...