The Q adopts a worse case scenario, i.e. commissions, slippage. And you can replace it of course, but replacing it negates your participation in any of the contests. But if you think about it, for a fund, the costs are not just commish and slippage, there are management fees and profit skimming. If you can make $ in simulation with such a heavy thumb on the scale then you're much more likely to actually make money trading in reality. Once again, per the contests, their game : their rules.

All the same, the commission models are not an accurate reflection of what's actually charged by the broker that Quantopian runs on. I understand the whole their game, their rules idea well enough. It just seems to me that if you wanted to build an online algo trading platform, you'd be interested in setting up your models to be as close to reality as possible.

I can think of two other distortions that come in to play.

Algo's that trade high priced equities have an advantage over those that trade less expensive equities. It cost 10x as much to buy an equal position in a stock quoted at $200 versus one at $20. Example: Buy $20,000 of a $200 stock and a $20 stock. Commission on that $200 stock is (20,000/200)*.03=$30. The $20 stock's commission is (20,000/20)*.03 = $300 . Thats a big difference and discourages algo's from that market corner.

The second impact is that since there is no minimum charge for commission, if you want to break up your $20,000 order in to 20 orders and issue one order a bar, you can minimize the slippage that is due to order size. (In approximate terms, the Q's slippage model is proportional to the square of the order size subject to some limits). Yet, your commission is the same for 20 small orders as it would be for one big order.

I agree with the original poster and Richard on this. Its a mystery where the "default commission" amount came from and it does both warp the type of stocks that are purchased and make otherwise good algorithms unusable if they need to buy and sell a bunch of stocks with lower ticker prices, which is completely arbitrary given there is little to no connection between stock price and market cap. While we're at it, the slippage model is also horrible compared to the real world. Just because a stock doesn't trade a bunch of shares in minute X doesn't mean that it doesn't have a large volume on the bid/ask at that time, all of which is tradable. Even a very liquid stock can have a few minutes during the day when it doesn't happen to trade a lot, and if your algorithm picks that minute to put in an order you take a huge hit with the slippage model. Even worse this doesn't ever even out, you can only take a hit with the slippage model, never benefit. Combining the issues with both the commission and slippage models has excluded several very promising algorithms of mine from the contest. In my case I'm taking them to live trading instead to make my point, but not everyone will have that option.
To Market Techs point, the management fees and profit skimming are reflected in how much of the fees and carry that they get are passed on to the managers, not reflected in commission and slippage which one would obviously assume would reflect "commission" and "slippage" respectively. If Qs need to cover their management fees and profit skimming was going to be passed on to contest winners then the contest winners would have to pay Q no matter what, given that Qs burn rate has to be over $100,000/6 months! It is Qs contest and their rules, however they are actively soliciting our input and very responsive to it, so it doesn't make sense for us to stop providing this input.

I think we've all lost track of why the Q even exists. It is my understanding that this contest is not it. This contest and the fund in general are hopefully profitable spin-offs from a SaaS style algo hosting platform. With that in mind, the built in market models can be replaced, at will, with custom designed versions that might more accurately model the markets one trades. So those clients who come here to trade their IB accounts (main LOB here) can provide their own market models. Note that I'm not going to defend Q's default market model, yeah it's way out of range in my opinion too; I was trying to pick possible explanations out of thin air with regards to fees and profit taking. But that said, what does it matter as far as their core business model is concerned? They're a service company. They want to get paid for providing a great service. Allowing custom market models to be swapped in is part of that service.

There's a few different questions in the thread.

• Why did Quantopian choose the default commission that it did? The simple answer is, it made sense at the time! We weren't sure we were going to build live trading, let alone what broker would be the first if we did. The numbers were the best-guess based on some industry conversations.
  
• Why hasn't Quantopian changed the default commission to reflect IB's pricing? This is a meaty question. Since it is configurable, for a long time the default didn't matter too much - algo writers could modify it to what they thought was a better number. The reason not to do it is consistency - we hate making old backtests invalid. But you're correctly pointing out that contest is different, because it's not configurable. I've just put updating the default on the list for closer review in upcoming contest versions.
  
• Why is the default slippage model set the way that it is? Isn't it too hard? This one is just a matter of conservatism, I think. We'd rather have an algorithm perform better in real $$$ than backtest than the inverse. We think that having a hard backtest is the safest way to move from test to real $$$. Also, it is possible to soften the slippage effects through more aggressive order management.
  
• What should the slippage model be? It's not an easy problem. A slippage model that has too little impact is a bad thing. A slippage model that has too much impact is a bad thing. The current version is intended to be an approximation that, when it errs, errs on the side of being too conservative. Making a slippage model that manages all trading volumes in all market conditions - one could spend a lifetime building that, and there is diminishing returns on the time invested. So if we should invest more in iterating on the default slippage model, what should that look like? How much better would it get? How much effort would it be for that improvement? Those aren't entirely rhetorical questions =)

And a couple comments:

• To be clear, when the contest winner's algo trades real money (when anyone trades real money on Quantopian), they get the real commission costs and real fill prices. It's not the case that we charge the contest winner $.03/share and pocket the difference. The algo winner is getting the true commission cost, no adjustment or filter. The information that the real money trader sees in the Quantopian dashboard is the exact same as what is seen in the IB dashboard.
  
• Quantopian's main business is the hedge fund, not the SaaS algo-hosting platform. We certainly have considered monthly fees for real-money trading as a business model, but we're convinced that the hedge fund model is a better one for us. When we announced the hedge fund in October we were announcing our overall company goal, not a sidelight.
  

Quantopian's main business is the hedge fund, not the SaaS algo-hosting platform.

Thanks for that clarification. The Q's original raison d'etre was that it was going to proletariatize the automated algorithmic trading world -- was it not? Take algo trading out of the hands of the elites and offer it to all traders regardless of skill or funding, no?

I must have missed this directional change. You guys truly are just mining the minds of starving quants then. I mean, it's probably going to be lucrative, at least you're betting it is. It's just that now your goals are not so noble. More along the lines of any avaricious hedge fund. No doubt you guys analyzed the space and found that there was no way to make money in a low margin trading SaaS. Ah well, par for the course in this sector.

Hi Dan,

Yes, thanks for the clarification, "When we announced the hedge fund in October we were announcing our overall company goal, not a sidelight." I'd figured the fund idea was a bit of an experiment, and not your main business plan. I see now on https://www.quantopian.com/about you say, without ambiguity:

Quantopian is a crowd-sourced hedge fund.

and,

Live trading algorithms can become part of our crowd-sourced hedge fund where top quant talent is matched with outside investor capital.

Grant

The good news is that perhaps the hedge fund will finance free live-money trading accounts for everyone that wants to do so, at the risk of being spotted by Thomas's Bayesian talent scout!

I appreciate Quantopian, the current plans as they have evolved seem to me a win-win for anyone with the interest and some coding skills. @MT, you can trade live/real without taking part in the hedge fund and make money.

Gary, oh I agree. The whole platform is a boon to quants everywhere. It just felt a bit like they turned to the Darkside with the formal declaration that they're now primarily a fund. You'll have to realize that ever since I started down the financial software road I've been conflicted. The whole industry remains as something that Vader would love to reign supreme over. And the worst part is that many participants (those that I've dealt with and read about) would welcome him to the job.

The Q's original raison d'etre was that it was going to proletariatize the automated algorithmic trading world -- was it not?

That was, and still is, our plan. We are still a platform which anyone can use to design, implement, test, and deploy trading algorithms. At least as of now -- I wouldn't presume to try to predict the future -- our intention is for that service to remain free. As Gary has pointed out, you can trade your algorithm and make money without participating in our fund, and as Simon has pointed out, it is to our advantage to keep the barriers to entry for our service as low as possible because the more quants there are, the more good ones there will be for us to invite to participate in the fund.

We are providing our members with things they can't get anywhere else, including our platform; the ability to build a track record of success before getting a job at a bank or hedge fund; and the possibility of being invited to participate in our fund and make more money than they would make just trading their own money.

We will continue to enhance our platform, adding new features, new tradeable instruments, new brokers, etc. Some of these enhancements, we will probably have to charge money for, simply because it will be too expensive for us not to. But many of them enhancements will be free. When we give our members things that benefit them, both we and our membes benefit.

It just felt a bit like they turned to the Darkside with the formal declaration that they're now primarily a fund.

Why? Investing in the stock market isn't evil. Making money from investing in the stock market isn't evil.

We're a for-profit business, not a charity. We want to change the world of quantitative investment -- heck, we already have -- but we also want to make money. We have to, to be able to continue providing the service to our members.

We have chosen the fund model because it has the greatest likelihood of success and the greatest potential upside both for us and for our members.

Quantopian and IB is a perfect match. I've used many other brokers in the past...IB is the best. Frankly, I think it is a distraction for Q to worry too much about adding more brokers. IB is very financially stable, is low cost, doesn't have any crazy exposures and overall is probably one of the safest brokers out there.
I've been live trading for almost a year with Q / IB and everything has worked 100% perfect. With Q, it has been the coolest, greatest financial services experience I've ever had.
Q should be very proud to have created such an exceptional experience and service for quants and investors.
Based on everything I've experienced, I have 100% confidence in Q that they will create an exceptional experience for investors in their fund.
Looking forward to learning when the fund is planned for launch.

Jonathan, I didn't mean to ruffle your feathers. I have a great time playing around on Q's site, picking and poking, it keeps my Alzheimers at bay.

Investing -- evil? No, not so much. Trading? I'm not so sure. In my early days in this business I equated traders to vampires: they design nothing of real social worth, create nothing of real social worth, produce nothing of real social worth and essentially can sit in the dark, working their mojo, grifting a living from the friction of the financial system. They are effectively sucking the life blood from actual investors through their machinations. Evil? Probably.

But lately I've taken on another less radicalized notion: traders provide the grease for the investment system. Even HFT, evil as the media makes it, seems to aid in the smooth execution of massive numbers of transactions. So, as socially distasteful as traders are (myself once included), they do provide lubrication to a multi-trillion dollar machine. A machine that probably would not work quite as well, quite so efficiently, without them.

Hmm.
Yeah I look forward to the moment of decision to maybe take half of some awesome profits and tackle worthwhile projects for the world with it (that's why I'm here), and then the other side (the one with the pitchfork) says, wait a second, you're going to be pulling $200k that at this rate could be $500k in two years, and see what I do, seems that decision turns some otherwise entrepreneurs into pure capitalists (using money only to make more money), and also even on the smaller scale a tough dynamic always.

@Simon,

The good news is that perhaps the hedge fund will finance free live-money trading accounts for everyone that wants to do so, at the risk of being spotted by Thomas's Bayesian talent scout!

Yeah, additionally one would hope that Q could work a deal with IB to pool the overall business being driven to them (although at this point, perhaps it is not yet significant), so that special rates would be applied.

Regarding scouting for talent, I think that Q has to be careful in this area. There is an automatic opt-in ( see https://www.quantopian.com/policies/privacy ):

We collect and store performance data generated by your algorithms when you use the platform to backtest, paper trade, and live trade.

Examining code is off-limits, but it appears that everything else is fair game. I guess it's part of the deal...free (or reduced-cost) resources and potential access to outside capital in exchange for access to algo performance results. It just seems like dicey territory (e.g. For the top performers, wouldn't it be interesting to have a peak at which securities are being bought/sold, when, and in what fashion?).

Hi Dan,

Regarding your comments under "What should the slippage model be?" you might consider the potential payback in developing a more accurate slippage model, unique to every algo (and similarly for the commission model). If I'm reading Jess' comment correctly on https://www.quantopian.com/posts/quantopian-managers-program-algorithm-selection-and-compensation, "...we have decided to open up our fund selection process to algorithms with "paper" track records..." it implies that if the model is off, you could end up making poor investment decisions based on trade simulations. Even if the model is off just a little, it could end up costing a lot in the long run (either because you wrongly filter out good algos or because you pick bad ones that appear to be good). You could cover the risk by gradually funding each algo (i.e. field test), but you'd end up dragging out the process to get your fund up and running, and incur higher development costs. So, I wouldn't underestimate the value in improving the accuracy of your market model.

Grant