Testing collective2

[schizo]

C2 algos can either list "magic fills" (based on shitty simulation) or by actual fills from subscribers, subject to the algo publisher's discretion (and practically, whether there are any subscribers).

Well, as you can see from below, the two don't add up.

From C2

I wasn't sure if it's based on Chicago or New York time, so I've included both charts for comparison. You will notice that the entry price lies outside the range (high and low) of both bars.

 

[Bad_Badness]

Was a while ago I spent any time on that site, but assuming no changes since I last checked:

C2 algos can either list "magic fills" (based on shitty simulation) or by actual fills from subscribers, subject to the algo publisher's discretion (and practically, whether there are any subscribers). The latter is time variant on market impact due to actual capital allocated to algorithm based on the subscribers. The order subscribers are filled is random. There are delays, not suitable for anything but high latency trading. If algo needs a market fill (for an adequately liquid instrument) a particular second (or possibly even minute) it's fudged.

I ran a single contract ES strat for 2 weeks on paper and then tried an actual account. Day 1: 2 errors. Day 2: 4 errors, Day 3: 7 errors. Errors were things like fills that were anywhere from ~12 seconds to ~4 minutes late. But the killer was the orders were so late, I already generated close orders. Then all sorts of chaos ensued: Positions closed, position double opened (overfills), Positions double closed (overfill and reversed). The syncc engine has to make a judgement on when it is out of syncc and WHEN to get back into syncc. You cannot try to out smart it because it is a black box.

Needless to say, I pulled the plug after that. I chatted with the guys at C2 and they were very helpful and transparent. Like Snuskpelle says, it works for swing type trading but not day trading where you hold for less than 10-15 minutes.

So regardless of the Sim or any other data, when the rubber hits the road and you actually trade, then you will uncover all the limitations.

 

[guest_trader_1]

You're right. The prices on the chart don't line up with the timestamps on C2. But what timezone is that timestamp? Chicago time?

The data provider (dukascopy) set timestamps as "local", so I assume they will be the ones at Chicago. They match almost perfectly with Tradingview. I double checked with them as well before posting anything.

 

[Radiohead]

Did you try reaching out to C2 to see what they say?

 

[Radiohead]

I ran a single contract ES strat for 2 weeks on paper and then tried an actual account. Day 1: 2 errors. Day 2: 4 errors, Day 3: 7 errors. Errors were things like fills that were anywhere from ~12 seconds to ~4 minutes late. But the killer was the orders were so late, I already generated close orders. Then all sorts of chaos ensued: Positions closed, position double opened (overfills), Positions double closed (overfill and reversed). The syncc engine has to make a judgement on when it is out of syncc and WHEN to get back into syncc. You cannot try to out smart it because it is a black box.

Needless to say, I pulled the plug after that. I chatted with the guys at C2 and they were very helpful and transparent. Like Snuskpelle says, it works for swing type trading but not day trading where you hold for less than 10-15 minutes.

So regardless of the Sim or any other data, when the rubber hits the road and you actually trade, then you will uncover all the limitations.

I agree, I wouldn't subscribe to anything on C2 that day trades. Swing trades fine. Orders that execute on market open and are known ahead of time, that's your best bet.

 

[lindq]

It is normal C2 trade fills are delayed.
In most cases it is broker's fault.
The print on C2 web is on right time, but subscriber's broker account get late fills. Then you will get an email telling you the fill on web is invalid, it needs to match whatever fill in subscriber's account.

Bad fills can also be generated by lazy subscribers who are trading manually. My last experience a few years ago involved a subscriber who logged into the system about an hour after I had triggered an entry. That caused a winning trade to be posted as a loss.

Any operation like C2 is going to have sand in the gears, through no fault of their own. It's the nature of dealing with multiple streams of data running both ways.

 

[guest_trader_1]

Bad fills can also be generated by lazy subscribers who are trading manually. My last experience a few years ago involved a subscriber who logged into the system about an hour after I had triggered an entry. That caused a winning trade to be posted as a loss.

Any operation like C2 is going to have sand in the gears, through no fault of their own. It's the nature of dealing with multiple streams of data running both ways.

Well,

- Strategies that don't match real market data, hence made-up results
- Delayed executions
- Unexpected errors

So much for "not fault of their own". I wrote the initial post to challenge a data set, but it seems that there is much more than simple made-up reports.