Best way to spot a new market

[zoltanthemage]

Sorry, I probably misplaced the thread on the "General betting" forum, that can be deleted/closed (http://forum.bdp.betfair.com/showthread.php?t=2949).

---

Hi everyone,

I am trying to find a way to be the first one (or nearly) to place a bet on a market, as soon as it is created, but I often find some other money before mine.

I have tried 2 ways:

- requesting in a cycle the list of events, as soon as I see a new event, I request the market catalogue for that event ID and I get the necessary info (market IDs and market types) to request the detailed markets books and place orders;

- requesting in a cycle the market catalogue for the market types I'm interested in (less than 1000), as soon as I see new markets I request the detailed market catalogue for those market IDs (I need the market type as the first request was for multiple types), then the market book and place orders.

The markets seem to appear first on the list of markets filtered by type, but then the detailed market description cannot be retrieved immediately (case 2).

In the end, it takes a lot of seconds to place an order, since I spotted the new event ID or market ID.

Any idea on the subject?
Thanks

---

Below the comparison between refreshing the list of events (case 1) and then requesting the market catalogue for the new events and refreshing the list of markets and then requesting the detailed market catalogue (case 2).

All times are UTC+2.

In the case 2:
- step 1 is a listMarketCatalogue filtering by the new market IDs and without runner description
- step 2 is a listMarketCatalogue filtering by the new market IDs with runner description
- step 3 is a listMarketCatalogue filtering by eventID and without runner description

The step 3 of the case 2 is the same request as the case 1, I only found out the new event ID in a different way. It's very weird to notice that the two processes have a non-empty response to that call at very different times (on the first example, the step 3 retrieves results at 14:11:42, while the case 1 at 14:12:04).


CASE 1:
2015-08-20 14:11:40 New event 27517723 added
2015-08-20 14:11:40 Checking market catalogue for the event 27517723
2015-08-20 14:11:54 Checking market catalogue for the event 27517723
2015-08-20 14:11:56 Checking market catalogue for the event 27517723
2015-08-20 14:11:58 Checking market catalogue for the event 27517723
2015-08-20 14:11:59 Checking market catalogue for the event 27517723
2015-08-20 14:12:04 Checking market catalogue for the event 27517723
2015-08-20 14:12:06 Running thread to place orders on the events 27517723 (types 7) and markets 1.120158240, 1.120158239, 1.120158243, 1.120158244, 1.120158247, 1.120158248, 1.120158252, 1.120158251, 1.120158255, 1.120158256, 1.120158260, 1.120158259, 1.120158263, 1.120158264

CASE 2:
2015-08-20 14:11:39 Step 2 success for 1.120158264, 1.120158263, 1.120158259, 1.120158260, 1.120158256, 1.120158255, 1.120158251, 1.120158252, 1.120158248, 1.120158247, 1.120158244, 1.120158243, 1.120158239, 1.120158240

2015-08-20 14:11:42 Step 1 success for 1.120158264, 1.120158263, 1.120158259, 1.120158260, 1.120158256, 1.120158255, 1.120158251, 1.120158252, 1.120158248, 1.120158247, 1.120158244, 1.120158243, 1.120158239, 1.120158240

2015-08-20 14:11:42 Step 3 success for 1.120158264, 1.120158263, 1.120158259, 1.120158260, 1.120158256, 1.120158255, 1.120158251, 1.120158252, 1.120158248, 1.120158247, 1.120158244, 1.120158243, 1.120158239, 1.120158240

---

CASE 1:
2015-08-20 14:25:54 New event 27517771 added
2015-08-20 14:25:59 Checking market catalogue for the event 27517771
2015-08-20 14:26:00 Running thread to place orders on the events 27517771 (types 7) and markets 1.120159964, 1.120159965, 1.120159968, 1.120159969, 1.120159972, 1.120159973, 1.120159976, 1.120159977, 1.120159980, 1.120159981, 1.120159985, 1.120159984, 1.120159989, 1.120159988

CASE 2:
2015-08-20 14:25:53 Step 2 success for 1.120159988, 1.120159989, 1.120159984, 1.120159985, 1.120159981, 1.120159980, 1.120159977, 1.120159976, 1.120159973, 1.120159972, 1.120159969, 1.120159968, 1.120159965, 1.120159964

2015-08-20 14:25:54 Step 1 success for 1.120159988, 1.120159989, 1.120159984, 1.120159985, 1.120159981, 1.120159980, 1.120159977, 1.120159976, 1.120159973, 1.120159972, 1.120159969, 1.120159968, 1.120159965, 1.120159964

2015-08-20 14:25:57 Step 3 success for 1.120159988, 1.120159989, 1.120159984, 1.120159985, 1.120159981, 1.120159980, 1.120159977, 1.120159976, 1.120159973, 1.120159972, 1.120159969, 1.120159968, 1.120159965, 1.120159964

---

CASE 1:
2015-08-20 14:26:56 New event 27517750 added
2015-08-20 14:26:56 Checking market catalogue for the event 27517750
2015-08-20 14:26:59 Checking market catalogue for the event 27517750
2015-08-20 14:27:02 Checking market catalogue for the event 27517750
2015-08-20 14:27:17 Checking market catalogue for the event 27517750
2015-08-20 14:27:22 Running thread to place orders on the events 27517750 (types 7) and markets 1.120159772, 1.120159771, 1.120159776, 1.120159775, 1.120159779, 1.120159780, 1.120159783, 1.120159784, 1.120159787, 1.120159788, 1.120159791, 1.120159792, 1.120159796, 1.120159795

CASE 2:
2015-08-20 14:26:59 Step 3 success for 1.120159795, 1.120159796, 1.120159792, 1.120159791, 1.120159788, 1.120159787, 1.120159784, 1.120159783, 1.120159780, 1.120159779, 1.120159775, 1.120159776, 1.120159771, 1.120159772

2015-08-20 14:27:00 Step 2 success for 1.120159795, 1.120159796, 1.120159792, 1.120159791, 1.120159788, 1.120159787, 1.120159784, 1.120159783, 1.120159780, 1.120159779, 1.120159775, 1.120159776, 1.120159771, 1.120159772

2015-08-20 14:27:03 Step 1 success for 1.120159795, 1.120159796, 1.120159792, 1.120159791, 1.120159788, 1.120159787, 1.120159784, 1.120159783, 1.120159780, 1.120159779, 1.120159775, 1.120159776, 1.120159771, 1.120159772

---

CASE 1:
2015-08-20 14:30:39 New event 27517774 added
2015-08-20 14:30:39 Checking market catalogue for the event 27517774
2015-08-20 14:30:41 Running thread to place orders on the events 27517774 (types 7) and markets 1.120160001, 1.120159999, 1.120160006, 1.120160005, 1.120160011, 1.120160012, 1.120160017, 1.120160016, 1.120160021, 1.120160023, 1.120160027, 1.120160028, 1.120160034, 1.120160037

CASE 2:
2015-08-20 14:30:38 Step 2 success for 1.120160037, 1.120160034, 1.120160028, 1.120160027, 1.120160023, 1.120160021, 1.120160016, 1.120160017, 1.120160012, 1.120160011, 1.120160005, 1.120160006, 1.120159999, 1.120160001

2015-08-20 14:30:39 Step 3 success for 1.120160037, 1.120160034, 1.120160028, 1.120160027, 1.120160023, 1.120160021, 1.120160016, 1.120160017, 1.120160012, 1.120160011, 1.120160005, 1.120160006, 1.120159999, 1.120160001

2015-08-20 14:30:40 Step 1 success for 1.120160037, 1.120160034, 1.120160028, 1.120160027, 1.120160023, 1.120160021, 1.120160016, 1.120160017, 1.120160012, 1.120160011, 1.120160005, 1.120160006, 1.120159999, 1.120160001

[Nick JD]

...but I often find some other money before mine.

I read somewhere a while back, mention that the bots in front of you might well be guessing market and selection ids to place these orders.

Not sure if this will discourage or encourage you, or if it is more than hearsay, but it does make some sense in the race to be at the front of market queues.

[zoltanthemage]

It seems a bit unlikely, but I can't say it's impossible. Would they at least request the market book? In that case I don't know if it responds non-empty before the market catalogue do. Otherwise I can't understand how they know what they're betting.

Anyway, my concern is finding the best "normal" method, also to limit the number of requests. I outlined two of them and, even if they should be equivalent, they sometimes show a big difference.

I'd like to understand which is the request to show a market first and why the same request from different processes does not display the new results at the same time. There might be a way to drive the desired behaviour, as it's usually the same kind of process to see the market later in my tests.

[jptrader]

It seems a bit unlikely, but I can't say it's impossible. Would they at least request the market book? In that case I don't know if it responds non-empty before the market catalogue do. Otherwise I can't understand how they know what they're betting.

Most of the people racing to get their bets on the market first are just laying at 1.01 anyway so they don't care what they're betting

[zoltanthemage]

Most of the people racing to get their bets on the market first are just laying at 1.01 anyway so they don't care what they're betting

Not all the markets are good for that, you have to at least restrict by eventType and if you guess a marketId, you have to at least request the market book and try to understand from there if it's what you are looking for.

Anyway, I understand what you're saying, I'm disregarding some not necessary info to save time now. But some checks need to be done, unless there is a logic connecting the market IDs to the market or event types, so it's not a wild but an educated guess.

[zoltanthemage]

With the above method, I've spotted the Andy Murray and Roger Federer next markets (i.e. 1.120387772, 1.120388399) when there were like 180k placed on the 1.01.

There is clearly something wrong, the fastest bots don't usually place that amount. In this case I must think there were users bets too if the amount is so huge.

The only other option is that there are a few fast bots that placed 30k-50k bets on that market. I think I noticed one that is placing 15k bets on some markets

Given that the choice of competition, market type and runners is accurate, I don't think the market ID can be guessed, unless the market book is available before the market catalogue and you can retrieve the necessary info from the market description.

[Nick JD]

I'm new to this stuff, but from what I've read up on: the golden age of the exchange has long since past. I hear of an exchange where there were numerous holes that systems could exploit to make free money.

The entropy in the exchange is now far too powerful for any "tricks" to be workable for anyone who hasn't got LOADS of cash, or a server in Malta exploiting high frequency trading principles.

A bit like arbitrage opportunities, if you're looking for a "trick" to profit, I get the feeling that that ship's sailed, or your system life-span will be annoyingly brief.

Technical methods to "predict the future correctly a percent or two more than incorrectly" may be the only methods left to gain an edge in this exchange ... much like the securities exchanges for the average chap.

I do believe there's room in this exchange for a decently thought out bot and a little bit of profit over the long term, but every bot writer retiring in 2017 are at odds that probably won't be matched, if you pardon my hideous pun.

2c

[zoltanthemage]

That's an interesting point of view, thank you very much for sharing it.

I'm also new to the business, so I'm trying to explore any possible way to build a successful betting bot, not only "tricks", and learn about them

I've worked for a couple of years on videogame bots where responsiveness and "being there first" were the keys, so the above "trick" has naturally been my first applied algorithm.

Would you explain why loads of cash would bring an advantage? In a race context like the above, I can't see why, but I might be missing something ofc.