Algo/Quant Traders (Interactive Brokers API) Performance/ Fund Raising

[TradeHippo]

(The above Chart S&P500 appears flat because of scaling since its cumulative returns is 94.83%)

Hi all anyone using IB API here?
Any Algo/Quant traders around?

Wondering if anyone manages fund for their friends/family here?

The above is backtest results, looking to test with my own money before branching out to manage for friends/family.

Trading only in US market for the above and benchmark it with S&P 500 (Bloomberg Ticker: SPY US Equity)

More than 2 times volatility of S&P 500 though.
Long/Short (Index/ Bonds/ Volatility)
23% Correlation with S&P
No Leverage
Max drawdown calculated monthly not daily

 

[Mecisteus]

Why not backtest from year 200X ?

 

[TradeHippo]

Why not backtest from year 200X ?

Total notional daily volume of all active VIX Futures contracts did not exceed $1 Billion until 2011 and as such you should take any VIX pricing data prior to 2011 with a grain of salt. Part of my algo is on volatility, So I wanted more data accurancy, but I have backtested some possible factors in my algo back until 20+ years too, those that are dependent on index and bonds..

 

[Mecisteus]

So you are saying if the VIX Futures contract did not exceed 1B, then the data is meaningless? Even $700M is not significant enough?

2010 till now is a stock/bond bull market only. Your algorithm must survive at least 1 bear and bull market cycle. Then you can consider to have found the holy grail.

 

[OngHuatHuat]

Can sell this for 1 billion usd?

 

[TradeHippo]

So you are saying if the VIX Futures contract did not exceed 1B, then the data is meaningless? Even $700M is not significant enough?

2010 till now is a stock/bond bull market only. Your algorithm must survive at least 1 bear and bull market cycle. Then you can consider to have found the holy grail.

Hi Mike thats right regarding the 2010 bull for stocks and bonds, thats my worry too, I expect the returns to fall to about 50% per year because I am worried that my data is over fitted to the past few years so I am researching on more longer dated data for indices and bonds ..... For VIX futures the further back it is from 2011 the less the value traded and it may add more inaccurancy..the further back the dates the more the value traded decline.. also I am Long/ Short so I am not worried about a bull/bear.. more on the data overfitting


Not considering this the holy grail, if you see my annualised vol its about 2.38 times the S&P... and if I am short vol and the market takes a sharp hit, i can be down 20-30% easily

 

[TradeHippo]

Can sell this for 1 billion usd?

No la.. who wants to buy, people can hire a few PHD guys and a few programmers and with enough research can develop a good strategy too... wont cost them more than 500k to do that

 

[wahkao3]

amazing retrns!

 

[PostCountWarrior[+1]]

how many trades a day? if in out many times, commission will add up.

did u take into account brokerage and bid offer spreads etc?

 

[TradeHippo]

how many trades a day? if in out many times, commission will add up.

did u take into account brokerage and bid offer spreads etc?

Hi PCW, its not a high frequency algo, it holds position from 1day to 10days approximately, thus commission is less of an impact, I factored in comms but did not factor in slippage, the reason being the slippage could be in or out of my favour, meaning I can sometimes only be queuing on the offer/bid, thereby earning the spread instead of at market where I will lose the spread, I expect it to be neutral in this case so didnt factor in.. also its all very liquid assets so the slippage is much lesser since high liquidity if there is any slippage.

 

[rayzzzz82]

How is your maximum DD calculated on? by taking the monthly DD, it does not reflect correctly the max risk you are subjected to. Considering the market volatility, it would be better to pep the maximum DD to the floating DD.

How reliable is your backtest data? It would be better to test it out from 2000 onwards in a full market cycle. from 2011 onwards, the US market is basically in a uptrend and it probably explain to why you have a great performance curve since 2011.

it would be good if you can live test for a least a year before recommending to your friends or anyone here. backtest is often not a reliable test. many can have marvellous backtest result but when it come to real live fwd testing, many other factor come into play and it will greatly affect the performance.

 

[TradeHippo]

How is your maximum DD calculated on? by taking the monthly DD, it does not reflect correctly the max risk you are subjected to. Considering the market volatility, it would be better to pep the maximum DD to the floating DD.

How reliable is your backtest data? It would be better to test it out from 2000 onwards in a full market cycle. from 2011 onwards, the US market is basically in a uptrend and it probably explain to why you have a great performance curve since 2011.

it would be good if you can live test for a least a year before recommending to your friends or anyone here. backtest is often not a reliable test. many can have marvellous backtest result but when it come to real live fwd testing, many other factor come into play and it will greatly affect the performance.

Hi Rayz, DD is calculated monthly, in this case its due to 2015 Jan which is the Full Max DD -10.4%, I am displaying full Max DD in this case over the 5yr period.

Test on bonds and index can be conducted >10years but not volatility as per earlier post because of less reliable data.

Regards to the uptrend, note that mine is a long/short, its not a long only, so not too biased towards the uptrend... it has only 23% correlation with the S&P 500

Yes definitely will go live myself first before my friends. Lets say that I should tell them to lower their expectations by 20%, with a disclaimer that historical performance does not necessary represent future performance.

 

[wahkao3]

backtest is often not a reliable test. many can have marvellous backtest result but when it come to real live fwd testing, many other factor come into play and it will greatly affect the performance.

if back test is not reliable, then what is reliable? any recommendation?

 

[wahkao3]

can you do a monte carlo test?

 

[TradeHippo]

I am not sure if a monte carlo test will be good representation as it uses a probability curve, but however you cannot factor in black swan events like Sept 11.. its more of data overfitting for recent years that I am more concern about..

 

[TradeHippo]

if back test is not reliable, then what is reliable? any recommendation?

All funds always come with a disclaimer: past performance does not necessary represent future performance

But to me I believe the market is not a random walk

 

[Mecisteus]

Short term is random walk.

The trend is clearer in the longer term for stocks.

 

[wahkao3]

I am not sure if a monte carlo test will be good representation as it uses a probability curve, but however you cannot factor in black swan events like Sept 11.. its more of data overfitting for recent years that I am more concern about..

monte carlo is ok lah. you need to model your returns as a probability curve.

 

[Shiny Things]

Couple of thoughts:

1) You'll want to go back and recalculate your max-DD numbers daily. A zero monthly DD isn't going to help you if IB stops you out halfway through the month because VIX blew up (I'm assuming you're shorting the VIX).
2) Just because there's no VIX futures data doesn't mean you can't backtest the strategy through 2008. See what happens if you sell SPX straddles instead of the VIX.
3) If your holding period is longer than a day (and yours is), I think you want to be taking liquidity instead of working orders and trying to earn the spread. If you're trying to capture a 3-5% move in whatever you're trading, then working an order to try to chisel an extra 0.01% seems like a false economy if it means you miss the trade completely.

 

[TradeHippo]

Couple of thoughts:

1) You'll want to go back and recalculate your max-DD numbers daily. A zero monthly DD isn't going to help you if IB stops you out halfway through the month because VIX blew up (I'm assuming you're shorting the VIX).
2) Just because there's no VIX futures data doesn't mean you can't backtest the strategy through 2008. See what happens if you sell SPX straddles instead of the VIX.
3) If your holding period is longer than a day (and yours is), I think you want to be taking liquidity instead of working orders and trying to earn the spread. If you're trying to capture a 3-5% move in whatever you're trading, then working an order to try to chisel an extra 0.01% seems like a false economy if it means you miss the trade completely.

Hi Shiny Things, thanks for the feedback

1) My monthly max DD is -10.4%, I dont have a zero monthly DD, 15% of my month is negative... I have calculated daily its about -17%

2) Good idea on the SPX straddles, hopefully I can find a way to retrieve this info.

3) We did some test on the spread, futures spread on less liquid futures can sometimes be wide and significant, so for example we queue on the bid/ask and if it did not get filled in say 1hr we cross the spread and get it done at market. So that we dont lose too much to slippage.