I don't understand

  • Thread starter Thread starter BuySellSideTrader2020
  • Start date Start date
rb7 said:
What was your backtesting period length?
Your sim period length?
Your live trading period length?
More...


backtesting is YTD
sim period length is a 3 months
live period is a week
 
Scataphagos said:
"Fill" shouldn't be the issue. If your results are significantly impacted by your fills, you're "fishin' in a dry hole" anyway.
...
The "type" of order will not make-or-break you, regardless.... that's "fractions on fills". Trades need to be better thought-out to be a big winner.
...
More...

Wow. That was ignorant. Definite "misses" from the Scat-man.

Bad fills are kind of like bad grounds in electronics.
"Back in the day," a bad ground might give you some static out of a speaker.
(The sort of robustness that was maxed (and typified) by the USSR using valve electronics in their fighter jets well after the US/NATO had gone solid state.)

But in 2018, any sort of ground issue in a sellfone or computer?? It's gone, with no recovery, of any component. Ka-PUT.

Fills in EOD systems can be a lot of things.
Fills in sub-hour, sub5-minute, sub-minute, and all the little fractions thereof, matter exponentially more.
 
Last edited:
trade4succes said:
In live mode, if you are waiting on the bid/offer to be filled, you are usually one of the last to go and also usually when you actually do not want to be filled.
More...
%%
REAL Liquid markets help; but i see your points+ agree with the first one.The last point maybe true in the last hour of trading; but entry for a mostly automated /swing positions with discretion, its fine.REAL liquid markets permit a market order; but i tend to use a generous limit order.- especially if no one hits my NICE price improvement........ LOL.I would not have any system dependant on market maker profits only; even though markets makers do fine with that.BSST2020 is not a market maker.:cool::cool:
 
BuySellSideTrader2020 said:
I have made a strategy that is extremely profitable on the simulated side yielding upwards of 10%-25% after commissions with a sharpe ratio between 7.00-10.00... when I run backtests, its comes out profitable, on simulation its profitable, but on the live side its barely profitable. I have accounted for slippage but I am unsure if my ping has anything to do with it, I have looked into low latency servers but want to make sure this is profitable before I spend the monthly fee.
I use the same number of contracts per entry as I do on the live side and still don't understand why one is profitable and the other is barely.
What do you guys suggest?
More...

I haven't really seen anyone mention this yet but a Sharpe Ratio of 7.00-10.00 is unbelievable. It would be my first sign something is amiss in my strategy. If this were true you could start your own hedge fund and basically beat the entire history of hedge funds up to the present day.

With that information, you know something is wrong already. A lot can be going wrong here:

  1. Are you doing any sort of walk-forward optimization? If not, it's very easy to deeply overfit your data and end up with poor live performance. This type of problem would be apparent quickly in walk-forward analysis of your strategy.
  2. Are you snooping? That is, are you using future data to determine the current trade? I'd place my bets on this being the most likely case given the presence of a Guinness Book of World Records setting Sharpe Ratio. I realize you can make the Sharpe Ratio arbitrarily high by reducing time frame but even then a 7.00-10.00 Sharpe on HF data is incredible.
  3. Fills are not the issue most likely. Even accounting for slippage in a market order you'd still have a very high Sharpe Ratio. You are trading the most liquid futures market.
  4. Trading on the one-minute bar is the territory of HFT. It's possible you determined a small strategy that worked on historical data but the present market microstructure has changed. This could be due to any number of reasons. There are entire books published on this.
 
gaussian said:
I haven't really seen anyone mention this yet but a Sharpe Ratio of 7.00-10.00 is unbelievable. It would be my first sign something is amiss in my strategy. If this were true you could start your own hedge fund and basically beat the entire history of hedge funds up to the present day.

With that information, you know something is wrong already. A lot can be going wrong here:

  1. Are you doing any sort of walk-forward optimization? If not, it's very easy to deeply overfit your data and end up with poor live performance. This type of problem would be apparent quickly in walk-forward analysis of your strategy.
  2. Are you snooping? That is, are you using future data to determine the current trade? I'd place my only on this being the most likely case given the presence of a Guinness Book of World Records setting Sharpe Ratio.
  3. Fills are not the issue most likely. Even accounting for slippage in a market order you'd still have a very high Sharpe Ratio.You are trading the most liquid futures market.
  4. Trading on the one-minute bar is the territory of HFT. It's possible you determined a small strategy that worked on historical data but the present market microstructure has changed. This could be due to any number of reasons. There are entire books published on this.
More...

I was shocked with the sharpe ratio as well, if it were to be 3-4 then I wouldn't be so skeptical. I am currently brainstorming ideas on improving the order fills.
 
You must log in or register to reply here.
Back
Top