Luck? or a good system?

pistolpt · May 2, 2012

Backtested a trading system that trades an index only three ways, 100% long, 0% (out of market) or 100% short from 2004-2011. This system returned 80% annualized with a 60% standard deviation over that period.

Are those GOOD results? Would you could consider this just data-fitting or curve-fitting?

The system is not that complicated (3 different signals) and there were at most 9 trades a month. Unfortunately there is no out of sample data I can test it on since the index was only trading 2004 onwards. In 2012 I am up 85% or so YTD.

abattia · May 2, 2012

Manual backtest, or automated backtest?

What is length of a typical trade (microseconds, seconds, minutes, hours, days, etc)?

"... trades an index ..." - did you backtest on index data, or future/ETF data? What timeframe of data did you use (tick, bars, end of day)?

Included commisions and slippage?

"...In 2012 I am up 85% or so YTD..." - is this a forward test (i.e. run in real-time with real market data)? Or is it live trading (in which case, well done!)?

bwolinsky · May 3, 2012

Quote from abattia:

"...In 2012 I am up 85% or so YTD..." - is this a forward test (i.e. run in real-time with real market data)? Or is it live trading (in which case, well done!)?

You know he's a noob talking about backtests.

braincell · May 3, 2012

They're most likely just luck. You need more info though, and a much larger sample.

Rationalize · May 3, 2012

Quote from pistolpt:

...Are those GOOD results? Would you could consider this just data-fitting or curve-fitting?...

...The system is not that complicated (3 different signals)...

I could probably answer better if I knew what the 3 signals were

Soon2Bgreat · May 3, 2012

VIX? IMO, a useful comparison would be against how buy/hold (or random long, short, out portfolio) over the same time period fared.

That said, regardless of risk/return, 60% vol seems high to me. Good to see it's paid off with live results though.

pistolpt · May 3, 2012

Quote from abattia:

Manual backtest, or automated backtest?

What is length of a typical trade (microseconds, seconds, minutes, hours, days, etc)?

"... trades an index ..." - did you backtest on index data, or future/ETF data? What timeframe of data did you use (tick, bars, end of day)?

Included commisions and slippage?

"...In 2012 I am up 85% or so YTD..." - is this a forward test (i.e. run in real-time with real market data)? Or is it live trading (in which case, well done!)?

-Automated backtest. In the sense that I came up with the criteria and then ran it against past data to see what the cumulative results would be.
(what would a manual backtest be?)
-Typical trade is at market close, once a day - buy & sell a couple VIX futures contracts each day (or some days no trading at all & 100% out of the market).
-I included commissions.
-Yes, that 85% is run with real time and real market data.

pistolpt · May 3, 2012

Quote from bwolinsky:

You know he's a noob talking about backtests.

I backtested it from Mar 2004 - Oct 2011. Been running it live since then.

But yes, I am a "noob". Gotta start somewhere man.

pistolpt · May 3, 2012

Quote from Soon2Bgreat:

VIX? IMO, a useful comparison would be against how buy/hold (or random long, short, out portfolio) over the same time period fared.

That said, regardless of risk/return, 60% vol seems high to me. Good to see it's paid off with live results though.

Ya, it's the VIX - VIX Futures to be exact. What do you mean buy/hold over the same period?

Yeah I was trying to find ways to reduce that vol. I mean, I worked out systems where I would trade 50% or less of the total portfolio, and pocket gains as I made then, but those results were more like 50% annualized return with 31% vol.
Working with options on those silly VIX ETN products now. Results were as I feared, lower returns with lower vol.

pistolpt · May 3, 2012

Quote from abattia:
Manual backtest, or automated backtest?
What is length of a typical trade (microseconds, seconds, minutes, hours, days, etc)?
"... trades an index ..." - did you backtest on index data, or future/ETF data? What timeframe of data did you use (tick, bars, end of day)?
Included commisions and slippage?
"...In 2012 I am up 85% or so YTD..." - is this a forward test (i.e. run in real-time with real market data)? Or is it live trading (in which case, well done!)?

-Automated Backtest - that is, I made the signals and then threw in past data to see how it would perform. (what is a manual backtest?)
-Trades futures contracts once per day, at market close. Just a handful. Sometimes no trades at all in a day.
-Backtested on Futures Data. End of Day.
-Included commisions.
-That 85% is live with real market data.