Does anyone use linear regression to day trade the markets?

[psytrade]

I backtested a long only system where the linRegSlope was > 0 on a 50% margin account on SP500 equities from 1965-2005 and the return was 27% and the drawdown was 63%.

I tried using it as a sell or short indicator linRegSlope < 0 but it didnt work out well.

This is for EOD trading. I dont know how it works for other timeframes

Statistics
All trades Long trades Short trades
Initial capital 100000.00 100000.00 100000.00
Ending capital 565021825.82 565021825.82 213662.46
Net Profit 564921825.82 564921825.82 113662.46
Net Profit % 564921.83 % 564921.83 % 113.66 %
Exposure % 82.80 % 82.80 % 0.00 %
Net Risk Adjusted Return % 682283.75 % 682283.75 % N/A
Annual Return % 27.23 % 27.23 % 2.14 %
Risk Adjusted Return % 32.88 % 32.88 % N/A

--------------------------------------------------------------------------------

All trades 7729 7729 (100.00 %) 0 (0.00 %)
Avg. Profit/Loss 72855.39 72855.39 N/A
Avg. Profit/Loss % 5.28 % 5.28 % N/A
Avg. Bars Held 44.04 44.04 N/A

--------------------------------------------------------------------------------

Winners 3295 (42.63 %) 3295 (42.63 %) 0 (0.00 %)
Total Profit 1729575899.14 1729575899.14 0.00
Avg. Profit 524909.23 524909.23 N/A
Avg. Profit % 17.88 % 17.88 % N/A
Avg. Bars Held 80.91 80.91 N/A
Max. Consecutive 31 31 0
Largest win 69331036.12 69331036.12 0.00
# bars in largest win 672 672 0

--------------------------------------------------------------------------------

Losers 4434 (57.37 %) 4434 (57.37 %) 0 (0.00 %)
Total Loss -1166476566.84 -1166476566.84 0.00
Avg. Loss -263075.45 -263075.45 N/A
Avg. Loss % -4.08 % -4.08 % N/A
Avg. Bars Held 16.64 16.64 N/A
Max. Consecutive 70 70 0
Largest loss -8269539.29 -8269539.29 0.00
# bars in largest loss 6 6 0

--------------------------------------------------------------------------------

Max. trade drawdown -64525433.34 -64525433.34 0.00
Max. trade % drawdown -66.52 % -66.52 % 0.00 %
Max. system drawdown -250614138.23 -250614138.23 0.00
Max. system % drawdown -63.99 % -63.99 % 0.00 %
Recovery Factor 2.25 2.25 N/A
CAR/MaxDD 0.43 0.43 N/A
RAR/MaxDD 0.51 0.51 N/A
Profit Factor 1.48 1.48 N/A
Payoff Ratio 2.00 2.00 N/A
Standard Error 94938250.42 94938250.42 4048.85
Risk-Reward Ratio 0.10 0.10 0.79
Ulcer Index 19.39 19.39 0.00
Ulcer Performance Index 1.13 1.13 N/A
Sharpe Ratio of trades 0.22 0.22 0.00
K-Ratio 0.01 0.01 0.08

 

[wint]

psytrade . . . I apologize but I can't make sense of any of that. I can't tell why you decided to go long or short. It sounds like you know what you're doing so I suspect it's my ignorance that's at fault. I never did much testing because I've found that most methods that are any good don't lend themselves to testing.

If I were to test, I would suggest something like this.

Find situations where trend is obviously up long term . . . say, 15 min chart, 9 pd sma > 18 pd, you've dran a linear regression line from lowest low to last high in current swing . . . price is inside 2std dev channel sloping up

Find pullbacks on 3 min chart; draw linear regression line on pullback

Enter on a Close outside channel top

As a safety measure use a 20 min ema on 3 min (or 34 ema which is equivalent of 20 ema on 5 min chart) and ensure price stays above it during correction

That's one of the simplest trading methods I know. If you can test that, I'm sure you'll be satisfied.

Nothing magic . . . it works because traders respect the 9 & 18, the upward slope channel, the 20 ema/34 ema.

You're just trading with the herd.