I have no idea if the following boast is true, but, how is this algo driven trading performance remotely possible ?:

“The founder of Tradebot, in Kansas City, Mo., told students in 2008 that his firm typically held stocks for 11 seconds. Tradebot, one of the biggest high-frequency traders around, had not had a losing day in four years, he said.

11 second average holding time, and never a losing day in 4 years? Sounds like bullshit to me . . .

NY Times via FT

Category: Quantitative, Trading

Please use the comments to demonstrate your own ignorance, unfamiliarity with empirical data and lack of respect for scientific knowledge. Be sure to create straw men and argue against things I have neither said nor implied. If you could repeat previously discredited memes or steer the conversation into irrelevant, off topic discussions, it would be appreciated. Lastly, kindly forgo all civility in your discourse . . . you are, after all, anonymous.

35 Responses to “Does Tradebot Have a Perfect Trading Record?”

  1. Cody Hill says:

    One of the TradeBot traders was my room mate for about two years. I’m pretty sure that should read 11 milliseconds or something similar. Their trades are next to instantaneous.

  2. hr says:

    It must be BS since they are doing better than GS.

  3. FrancoisT says:

    So, they invented the financial perpetual motion machine?

    Mbwahahahahahaha!

    Sounds like a Everest-high steaming pile of bacteria-laden bovine digestive remains.

  4. grunge101 says:

    Dipping my toes in this area recently, I don’t see that as impossible. If you stick with generally winning situations and aggregate them across many stocks and multiple strategies, you can likely achieve that.

    I expect it is the losing trades that lead to the 11 second average holding time rather than being typical for winning trades – then again, market making strategies could be very quick.

  5. george matkov says:

    Extraordinarily improbable, I agree. However, it is theoretically possible. He doesn’t claim no losing trades, rather no losing days. If you read Patterson’s excellent “The Quants” [well, you probably did know this already] you know that, in order to succeed, algo formulas must be kept secret. If Tradebot has found a short-term flaw in the markets it can exploit it until others figure out what he is doing and create an algo to beat it or someone at the firm has a big mouth and can’t help give out enough hints. They also have to stay small enough to remain under the radar.

    I’ll make a side prediction: some really smart quants are poring over the trading activity between an hour before the 15 min. meltdown/meltup last Thursday and maybe a hour after. Algos that were shut down as the market dropped and then restarted can’t help but to have left visible tracks from which their trading secrets can be extracted. If you read Stephenson’s Cryptonomicon you will recognize that this is a variant of how the Enigma machine’s secrets were eventually revealed.

    What say ye sir?

  6. EvilTraderDude says:

    Entirely possible. Manually scalping futures for a small profit is child’s play. Imagine what a machine could do!

  7. genomik says:

    The fact that there is a highly successful company like this with the name “Tradebots” itself should be reason for concern from humans. I have a friend who developed genetic algorithm trading bots 8 years ago. Each bot got a stocks and learned from trading mistakes. They got better and if they lost money too much they “died”. He did pretty good with it. That was just one small company, imagine what a company like Tradebots could do with lots of resources.

    All us humans out here should not scoff at the idea that this company could do this. Even if it was 80% true and they had profitable days 80% of the time (and minimal losses on bad days), that would be better than most/all humans.

    This company will no doubt get better and there will be competition from other domestic companies and from international. With Moores law, it will just get better.

    I believe that soon it may be the case that humans are not needed on Wall Street much. The data, technicals, fundamentals and news feeds are all digital and able to read by bots. Maybe traders jobs will be soon automated away. “All hail our robotic overlord, Tradebots” ;-)

  8. peterpeter says:

    Suppose Tradebot trades 100 million shares per day ( just to use a round number in the right order of magnitude). That would be 50 million shares traded round trip (assuming they close their books flat each day).

    If they are net of SEC, NASD, NSCC, ACT, and exchange fees/rebates seeing 51% of their trades be net profitable, then an average day would yield 1M shares traded at a profit.

    With those kinds of numbers (and even much slimmer win percentages much closer to 50%), it is statistically very hard to have a losing trading day without material errors in the system (bad feed data undetected, error in code etc).

    Given a large enough sample size and fast enough computers able to pull mis-priced limit orders before someone else hits them, spread betters will always make money. That said, each trade will have nearly a 50% chance (likely just under) of losing money, so each trade does carry risk.

    There is nothing nefarious about this. it is just the law of large numbers at work.

    Because the HFT market makers can calculate where they want their orders placed much more accurately and quickly than their human counter-parts, they are able to profit from razor thin margins at high volume and offer much tighter spreads, which helps everyone.

    I’d rather give Tradebot (or any of their hundreds of competitors) a 1c spread, than be gouged by a human market maker with spreads of an 1/8th of a dollar.

  9. mbelardes says:

    Wasn’t there some guy that had a similar trading record with an equally absurd boast about a trading algo that was called into question?

    What was his name? Madoff?

  10. taylorhr says:

    Slightly off topic….if a machine (program, algo, etc.) existed which did nothing but make money at will, wouldn’t that be the most powerful machine in the world (for it’s owners)? Think of the ramifications, I don’t think that humans are capable of the responsibility that would need to go with such a thing.

  11. peterpeter says:

    > mbelardes Says:
    > Wasn’t there some guy that had a similar trading record with an equally absurd boast about a trading
    > algo that was called into question?
    >
    > What was his name? Madoff?

    You don’t however see Tradebot, or any other successful HFT fund looking for outside capital, quite distinct from Madoff.

    HFT funds with low average holding periods for each position have very little need for capital (and are not highly leveraged), since the same $1M can be traded hundreds of times per day. On top of that, returns are typically much much higher than the 2% most hedge funds charge as a management fee… so successful HFT funds do not look for outside money.

    If someone asks you to invest in a HFT fund, run. It is either generating poor returns on existing capital, or a scam.

  12. dcsos says:

    Barrys gonna be angry because he works for his returns.
    These guys only hadda push stop on May 6th.
    Its very annoying to be shown that its rigged enough to fly a solution like Tradebot.
    Makes all the work you do….useless.

  13. Joey says:

    yes…….I think its possible.

    It doesn’t really take much for these guys to make a profit. From what I can tell, as an instrument nears meaningful levels of importance, they have ample opportunity to exploit the volatility and order flow. Being an intra-day trader, I see programs executing multiple time per day in one form or another.

    p.s. Your site is acting goofy. Very slow to load (if at all)

  14. b_thunder says:

    “…and never a losing day in 4 years? Sounds like bullshit to me” — not if they are FRONT-RUNNIG

  15. PhilB says:

    Here’s a question for you:

    Does Germany’s proposed ban on naked short-selling affect all these tradebots? If you short a stock for 11 seconds before you buy it back, is that considered a naked short-sell if stocks were secured for those 11 seconds?

    If so, and Germany’s proposal goes universal, this could have massive impact on HFT.

  16. GB says:

    When they turned off the liquidity machine and made the market “normal” again I think it shows how much they had been in control of the market. IMO. I can’t see how you could lose…

  17. peterpeter says:

    > “…and never a losing day in 4 years? Sounds like bullshit to me” — not if they are FRONT-RUNNIG

    Don’t conflate illegal front-running with legal order anticipation.

    You can’t front-run if you don’t have any customers.

  18. R. Cain says:

    in computer terms (3 GHz = 3 billion cycles/second)

    11 seconds is enough time to front run 10,000 trades, have lunch, and get the car washed

  19. Slightly off topic….if a machine (program, algo, etc.) existed which did nothing but make money at will, wouldn’t that be the most powerful machine in the world (for it’s owners)? T

    It depends on how much money it made. The world is in constant growth mode (inflationary) so only if it grew at a rate faster then inflation would it be able to break the bank. It could fool the owners into thinking it is making money but all it would be doing is matching the rate of inflation. As someone has succinctly put it, “you can only beat the market for so long before you become the market.”

  20. b_thunder says:

    peterpeter: what do you call a trading “universe” where HFTs can sniff out and “intercept” your order before it reaches the exchange and trade ahead of it? whatever you call it, that’s (imho) the reason for consistent profits that would make Madoff jealous!

  21. Nick Abe says:

    My view is that this is entirely possible. What people have not asked is how many absolute dollars are being made? It would seem relatively easy to buy/sell instantly and make a penny (or less) and just have your machines track momentum. The problem is you need to subtract data feed costs, salaries, and other fees.

    As someone above mentioned, if you trade 100mm shares/day and that’s 50mm round trip and you made on average (net of fees) 0.05c/share it’s something like $25,000/day. That’s good money, but not enough to build an empire on.

  22. peterpeter says:

    > b_thunder Says:
    >
    > peterpeter: what do you call a trading “universe” where HFTs can sniff out and “intercept” your order
    > before it reaches the exchange and trade ahead of it? whatever you call it, that’s (imho) the reason for
    > consistent profits that would make Madoff jealous!

    I call it science fiction!

    HFT can not intercept orders. Your broker can internalize your order before it hits an exchange, but since funds like Tradbot have no client orders, they don’t see any order flow before the quotes come out of the exchange feeds…. which anyone can subscribe to.

    Retail orders are 1-and-done, meaning that there is no way to anticipate order flow from looking at the first order, since there is no follow-on order (i.e. retail traders do not trade blocks of shares with dumb algorithms that can be anticipated).

    If a large fund is moving a block of shares in a way that exposes their intentions to other traders, I see absolutely nothing wrong trying to profit from that information, I fault the large fund for sloppy trading, and I still believe that the large fund has lower execution costs today than in years past (where human market makers would gouge them, and the commissions per share were much higher due to the lower overall volumes).

    Someone will now surely scream about flash orders… (and if you are using a broker that sends your orders to Direct Edge blindly whether or not you want to have your orders flashed, then I suggest switching brokers!), but Direct Edge in total (a superset of flashed orders) handles only 13% of the volume in US markets. Flash trading is a rallying cry for people upset at the huge profits at some funds (especially those institutions suckling from the government via TARP), but it is a red herring. Schumer was so gung ho about a ban on Flash trading so that he could get volume back to NYSE…

  23. wunsacon says:

    I believe this development is not just possible but inevitable.

    Actually, it’s what I would prefer to work on…

  24. Deferred Comp says:

    Dear Mr. Peter Peter,
    You have obviously never traded or executed institutional order flow. Your description of block trading is amateur. And those mythical lower execution costs… all made up for in one afternoon. “Legal order anticipation” is a very nice description of the message traffic mafia that is HFT. Any “gouging” by human market makers pales in comparison to what now passes as market structure. Optional market participation and fictional liquidity are bullshit. Knock yourself out defending it.

  25. daveirl says:

    Explain how high frequency traders front run orders if they don’t have an order to front run. The rights and wrongs of anticipating what’s going on in the market is a different issue but it’s clearly not front running. As peterpeter just mentioned if people have an issue with flash orders then don’t use them.

    There’s a massive misunderstanding of what front running is online. I actually trade institutional orders for a living, I’d love HFT to die and the market to move back to larger blocks on exchange but I don’t use phrases like front running to describe things that aren’t front running.

  26. onefive says:

    Considering the degree of success Tradebot has experienced since its “garage startup” roots in 1999, whether or not they have had a losing day in the last four years seems pretty irrelevant but also, entirely plausible. It’s perhaps worth noting that the founder of Tradebot also launched BATS in reaction to NASDAQ buying BRUT and INET. BATS is now the third largest exchange by share volume so it is probably safe to say they must have been doing something exceptionally right along the way.

  27. Hit the Reset button says:

    If you can see orders before they hit, you probably could have a high win percentage, don’t know about a 4 year win streak though, sounds more like trader ego after a few drinks…

    You know what they say “The best traders never talk about their trading methods”

  28. damo says:

    The more important point, I think, is that if these guys are so good a providing liquidity then why is their holding time only 11 seconds?

    Doesn’t this mean that if we didn’t have them I would have had to wait 11 more seconds before executing my order – and I might be a few tenths of a cent in front… so tell me again about how valuable this liquidity is that they provide?

  29. nemo says:

    The part about never a losing day for 4 years sounds like obvious puffery. You sure Tradebot founder David Cummings isn’t really Aleksey Vayner, Yale undergrad, finance genius, hedge fund CEO, world-class athlete, ballroom dancer, CIA assassin, and all around liar?

    http://en.wikipedia.org/wiki/Impossible_Is_Nothing_%28video_r%C3%A9sum%C3%A9%29

  30. Pir2 says:

    $0.01 a day means it is a up day. 0.01*365*4= $14.6.
    will be easy to check, give them $1000 and check in 2 minutes.

    where is the Jules Verne or the Philip k Dick of the trading fiction?

  31. mcknz says:

    Dykstra had a perfect record too. He just never realized any of his losses.

  32. dedalus says:

    Dear Mr. Deferred Comp:

    Do you think institutional order flow does NOT like trading with HFT firms?

    You tell peterpeter:

    “You have obviously never traded or executed institutional order flow. Your description of block trading is amateur….”

    Hmm. . . .

    Vanguard’s Chief Investment Officer George (Gus) Sauter oversees $920 billion in investments. Sauter says “We do think [high-frequency trading] enhances the marketplace for all traders.’ ”

    http://www.forbes.com/forbes/2009/0921/revolutionaries-stocks-getco-new-masters-of-wall-street_print.html

    Does Sauter’s belief that “[HFT] enhances the marketplace for all traders” signify that he too is an “amateur”?

  33. Karen2 says:

    Themis Trading, LLC has a wealth of material on HFT (High Frequency Trading), some of which may explain why Tradebot could have a perfect record of no losing days in 4 years.

    Their January 2009 white paper “Toxic Trading on Wall Street” lists several HFT strategies:
    http://blog.themistrading.com/?p=29

    And on May 11 (2010) they identifty two exchanges that feed enough details about “hidden” trades to their HFT customers to allow them a good chance of figuring out others’ trading strategies:
    http://blog.themistrading.com/?p=906

    From what I understand, HFT is basically about extracting a very small profit each from a lot of trades in which the HF trader takes advantage of superior computing speed to “jump in” ahead of a real buyer to buy from the seller only to sell almost immediately to the buyer they jumped in ahead of, at a slightly higher price than they paid. They basically take advantage of situations where a seller’s lowest price is a bit higher than a buyer’s highest price.

    Sounds leech-like to me. They’re charging other market participants a bid-ask spread without performing any service to benefit those market participants.

  34. Karen2 says:

    Oops – make that LOWER than a buyer’s highest price!

  35. Deferred Comp Says: May 18th, 2010 at 5:06 pm

    Dear Mr. Peter Peter,
    You have obviously never traded or executed institutional order flow. Your description of block trading is amateur. And those mythical lower execution costs… all made up for in one afternoon. “Legal order anticipation” is a very nice description of the message traffic mafia that is HFT. Any “gouging” by human market makers pales in comparison to what now passes as market structure. Optional market participation and fictional liquidity are bullshit. Knock yourself out defending it.
    ~~~
    this, above, is a tremendously interesting topic..People that are comparing “the ~1/8 spreads” of yesteryear, to “the ~.01 spreads” of today, should wonder..

    it may be stretching the point, but, it is similiar to being on the High-Diving board, looking into the Well, below..do you want “Water”, or “Cool Whip”, in that Well, before you Jump?
    ~~
    and, w. this: R. Cain Says: May 18th, 2010 at 2:40 pm

    in computer terms (3 GHz = 3 billion cycles/second)

    11 seconds is enough time to front run 10,000 trades, have lunch, and get the car washed”
    ~~
    is a good point that seems to illuminate the some of “Darkness” shrouding this topic..

    peep may do well understanding some of the componentry..
    http://clusty.com/search?input-form=clusty-simple&v%3Asources=webplus&query=genetic+algorithms
    http://clusty.com/search?input-form=clusty-simple&v%3Asources=webplus&query=neural+networks
    http://clusty.com/search?input-form=clusty-simple&v%3Asources=webplus&query=Informatics+Analytics
    for starters..and, may be useful to overhear some of those HFT ‘Jocks’ bitch about, and wanting discounts for, being in ‘the far Rack’ at the ‘co-location’ Center..