Re MCProTrader system

What do you people think of the MCProTrader system? The performance to drawdown ratio so far is fantastic. Though some of the trades don’t look quite right. There are option trades where the system bought an option for $0.05 and sold it for $0.10. Now $0.05 is normally the lowest price for a vix option. It looks like the system was buying at the bid and selling at the ask. This was before the system had any autotraders following, so there is no real fill data available for those trades.

Options are easy to play game.
With bid 0, ask 0.10, Someone can place a limit buy at 0.05 at C2, and place a limit sell at 0.05 at options exchange, then quickly cancel the order, his C2 order will be filled in a second.
I am not saying he was doing this.
I would watch if he can repeat the performance he did before subscribers come in.

So he sent a system broadcast in which he mentioned

Since the system opened to subscribers on 2/16/2017, it has gained 8%, so it is on track. The record so far is much better than 100% annual return. This is due to the fact that C2 handle orders differently before and after having subscribers. Before subscribers, an order will be filled if the limit price is touched. The spread is working for my favor.

So it seems he’s been ranked 3rd or 4th on C2 based on trades that would not be possible to do in a real account. The 8% return since system gained subscribers is quite good but not as strong as the pervious performance which was over 30% a month. The current C2 rank is misleading to subscribers.

You nailed it , using this method i can replicate his performance just in one day .

1 Like

PropTrader - I wanted to see how you would do it, but the system you are linking is not shown as it is marked private.

Opps i am sorry i set it to private and then killed the system , cant make it public now . Made 67% return with 0.6% max dd just in 24 hours . All using limit orders in Vix futures .

2 Likes

thanks for illustration.
So, if I get it right, the simulated limit orders on C2 are just not working correctly and lead to incorrect simulated performance numbers.
Thanks for the hint. I will watch out for this, when evaluating systems!

If the system didnt have any auto traders at the time .

If collective2 doesn’t have a way to reasonably estimate executions without subscribers, then all of the “track record” should not be presented, with or without disclaimers about it being theoretic.

And if you only have less then a month of real performance data, then it doesn’t matter if it is 2%, 8% or 20%.

Hi, Gal - I’ve asked PropTrader to explain to me via private email the mechanism he used to game the use of limit orders. Once I understand, I’ll close the loophole and take appropriate actions.

4 Likes

Hi MK i just replied to you by email . BTW i am not Mcprotrader :slight_smile:

To be fair to C2 i would like to add an important note : that by using limit orders there is no guarantee that the trade will end in the black , market can easily move against the system developer many points after buys getting filled at the bid - or sells getting filled at the ask - .

Okay!

PropTrader and I exchanged emails. What I wanted to confirm – the most important thing – is that there is not some secret “exploit” that allows someone to maliciously post a super-profitable strategy on C2 with 100% certainty that each trade is profitable, and no risk of failure.

That is not the case here. :slight_smile:

What PropTrader has described (to add more detail to his previous post) is simply this:

For strategies without AutoTraders, Collective2 assumes that – in the case of limit orders – you will get filled at a bid (for Sells) or at an ask (for Buys).

In real life, this doesn’t always happen, because orders that do not “trade through” the limit (i.e. go above or below them) may not have sufficient volume in the marketplace to fill all bids or asks that are posted. Yes, some orders will get filled, but it is not guaranteed that all will be filled, nor (more importantly) that all C2 Members’ orders will be filled.

(Just an aside: the fabulous chart PropTrader posted was indeed unrealistic in this specific regard, but it could have been unprofitable, too. His simulated trading enjoyed a bit of luck in that prices didn’t move down during the time he entered his trades. In other words, this is not a software exploit.)

The problems of limit orders that do not “trade through” the limit price is a well-known problem in simulated trading, and there is no perfect solution. After all, it’s conceivable that there might be enough volume to fill many or even all AutoTraders, even if the limit order does not trade through.

So, sure, C2 could absolutely require a trade-through to fill a limit-order in cases of no AutoTraders. However… Strategy Developers could make the reasonable argument: “Look, while it’s true that there is no guarantee that all people would be filled, it’s equally not true that all people will not be filled.”

That is why I really like C2’s use of real-life AutoTrading fills to determine the fill prices of strategies with AutoTraders. We actually show you how many orders really got filled, at what price, etc. The problem is solved automatically through the use of real-life fill data provided by real brokers from real accounts.

The challenge is: C2’s solution is only possible when a strategy has AutoTraders.

For now, what I suggest to members of Collective2 is to always remember C2’s admonition that results displayed on C2 are hypothetical. Even in cases where our data is based on real-life fill prices, remember that there is no single account that looks exactly like the C2 Model account.

More important than that: I think subscribers ought to strongly discount results that do not include AutoTrade data, particularly if the trades are limit trades and the profits on those trades represent are not much bigger than the bid/ask spread of the instrument.

This is yet one more reason to strongly encourage all Strategy Developers to actually trade their own strategy, either through the use of BrokerTransmit, or by AutoTrading your own strategy just like a subscriber would, in a real broker account (that is, if you do not use BrokerTransmit to generate your strategy’s orders).

Another possibility I am considering is somehow visually flagging on the C2 Strategy track record those trades that seem less likely, and when there is no AutoTrading data.

I should caution that this opens its own can of worms, and I’m not 100% sure what the regulators will think of this idea (does flagging a trade a “less likely” somehow imply that other simulated fills are “more likely” …yes, we always need to think about UI in these terms…)

But I will noodle this and see if there is some kosher way to visually flag “scalpy” limit-order simulated fills.

4 Likes

Didnt you have “keep after worst case slippage” in the past ? I recall you were comparing avg profit to avg bid/ask spread . Maybe bringing this back will solve “the issue” here .

Yes, that was in the right ball park. But I think I’d like something a bit more exact than that. Once there are real-life fills in C2, the slippage and spread costs are “baked into” the results. What I’m thinking about is some kind of analysis that discounts non-AutoTraded fills by looking at the P/L versus the spread. So, sort of like that stat you mentioned, but more selectively applied to sim-only data.

1 Like

I don’t see how you can realistically estimate performance in case of short term trades with such large spreads.

Looking at P/L vs. spread size can be a measure for trade quality, however it will probably penalize options trading estimated results in the same way that filling only if the price trades through the bid/ask will.

Perhaps sending the trades to an IB paper-trading account will produce more reliable results?

It seems to me from this discussion that McProtrader is being unfairly branded as someone who is gaming the system. Looking at the trade record, only two out of 44 trades look suspicious to me: one on 1/9/2017 and the one on 1/12/2017 that MaxTor is doubtless referring to, buying at $.05 and selling 4 minutes later at $.10. Another trade in December had the same spread but the trade had well over an hour to evolve … so it looks completely legitimate. Everything else looks like great trading to me. System performance always takes a bit of a hit when the auto-trading starts – nothing unusual there. Remember HFT VIX Scalper, one of Jonathan Kinlay’s systems from last year? Now that was a classic example of unachievable results.

@MarkAmspoker What you said is inaccurate , the discussion here applies to most of his trades , bear in mind that entry doesnt have to be 0.05 it could be 0.25 or 1.5 or any figure , what matters its clear that he used limit to buy limit to sell to execute his trades which most of the time is not executable in auto trading live accounts due to exchange filling algos , so i repeat it doesn’t have to be a one tick profit trade , it could be long at 0.5 exit at 0.65 , what matters here is limit orders were used to enter and exit the trades , bottom line his fills are "unmarketed " most of the time . Add to that that he admits thats the case here in his pm to subscribers as quoted below

I am trading this system and so far it’s doing very well. I am autotrading it.

The simplest solution is to add a column to Leader Board (and to the Grid) showing Annual/Cumulative return since Autotrading began. You can’t ignore the pre-Autotrading returns because that is the only method by which new Trade Leaders can get any subscriptions. And weighting pre- and post-Autotrading returns would be a nightmare.

Adding a column for returns since Autotrading began would be analogous to the practice of some mutual funds that show “returns since inception,” as well as “returns since open for investment.” Some mutual funds are started in private and then (if they do well) opened to the public. That is similar to what Providers do on Collective2.

If you did this, smart investors here would then give very little attention to returns BEFORE Autotrading began and let some other brave soul be the first subscriber to begin Autotrading on that strategy.

BTW, you might add more columns to the Leader Board anyway, for example, Last 3 months return and Last 12 months return (both without substituting cumulative returns for new strategies). That would also tend to give a better sense of how strategies are working after Autotrading begins (even though that data would not be tied necessarily to Autotrading). The fonts and graphics are large on the Leader Board and there would be room for more columns.

4 Likes