The betfair package: An R implementation of the Betfair APIpast.rinfinance.com/agenda/2011/BryanLewis.pdf · The betfair package: An R implementation of the Betfair API Bryan Lewis

The betfair package: An R implementation ofthe Betfair API

Bryan Lewis and Colin Magee

http://betwise.co.uk

R/Finance 2011

http://www.betwise.co.uk/smartform

Betfair

• Founded in 1999 by Andrew Black and Ed Wray.• Sports betting, casino games, poker, and more recently, FX

and CFD exchanges.• Processes more than five million transactions every day.• Three million betting exchange customers.


Betting APIs

Global sports exchange• Real time market data feed and clearing house for: Horse

racing, tennis, football (including American), basketball, andalmost anything else you can bet on...

Games exchange• Poker, casino games

Exchanges operate across globally distributed data centers.

The APIs are free to use! (higher-speed commercial versions arealso available).


The betfair Package

Implements the sports exchange API in R (depends on Rcurl∗).The API defines R functions in two broad categories:• Market data.• User account (place/update/cancel/monitor bets).

All functions use the https protocol and closely follow the syntax ofthe Betfair API documentation.

∗Rcurl is available for Windows fromhttp://www.stats.ox.ac.uk/pub/RWin/bin/windows/contrib/2.13/.


Package FunctionsAccount-related

login logout keepAlive placeBets cancelBetscancelBetsByMarket updateBets getBetgetCurrentBets getBetHistorygetMatchedandUnmatchedBetsgetMarketProfitAndLoss getAccountFundsgetAccountStatement viewProfile

Data-related

convertCurrency getActiveEventTypesgetAllCurrencies getAllEventTypes getAllMarketsgetCompleteMarketPricesCompressedgetDetailAvailableMarketDepth getEventsgetInPlayMarkets getMarket getMarketInfogetMarketPrices getMarketTradedVolumegetPrivateMarkets getSilks


Example Session

> library(’betfair’)> login(username=’******’, password=’********’)> getActiveEventTypes()[1:8,1:3]

id name nextMarketId6423 American Football 07511 Baseball 07522 Basketball 0

136332 Chess 04 Cricket 011 Cycling 03 Golf 013 Horse Racing - Todays Card 102707815


Market Info

> getMarketInfo(102707815)$marketStatus[1] "ACTIVE"$marketSuspendTime[1] "2011-04-27 18:43:00 GMT"$numberOfRunners[1] 9$delay[1] 0$reconciled[1] FALSE$openForBspBetting[1] "true"


Market Info

> getMarket(102707815)$runnersname selectionId

Monslewn 5503082Destiny Joy 5503083

Sweet Argument 5343101Dower 5503084

Napa Love 5503085Eager Leader 4914430Vivs Tiara 5503086

Daiseysgoneagain 4656431Rebas Affair 5503087Mi Dulce Koko 5503088


Order Book

x = getCompleteMarketPricesCompressed(102707815)plotPrice(x[’5503084’]); plotPrice(x[’4914430’])

To back current price 2.94

Total amount matched 169.5

30 25 20 15 10 5 0

2.52

2.56

2.58

2.64

2.72

2.74

2.88

2.9

2.92

2.94

3.25

3.35

3.5

3.6

6.8

7.2

7.8

8.4

8.8

11.5

To lay current price 3.25

Last price matched 2.92

0 5 10 15 20 25 30

To back current price 9.8

Total amount matched 0.28

14 10 8 6 4 2 0

5.1

5.5

5.7

6.4

6.6

8.4

9.2

9.4

9.6

9.8

30

36

38

55

240

To lay current price 30

Last price matched 38

0 2 4 6 8 10 12 14


Placing a Bet

bet = newBet(betType=’B’, marketId=102707815,price=2.9, selectionId=5503084, size=100)

b = placeBets(bet)

Note that bets may be partially matched. Interrogate bet statuswith:

gb = getBet(b$betId)gb$betStatus[1] "M"

gb$avgPrice[1] 2.90


Simple Time Series Example

Many R packages for quantitative finance apply to savedexchange data.

Betfair does not (yet) provide a historic data API. We saved thematched back prices every second for a race at Cheltenham inthis example:

library(quantmod)load(’chelt.rdata’)LocalHero = bp1[,"Local Hero"]chartSeries(LocalHero, theme=chartTheme(’white’))overround = as.xts(rowSums(1/bp1) - 1,

order.by=index(bp1))addTA(overround, col=2)addBBands()


Back Price for a Runner and Market Overround

2.6

2.8

3.0

3.2

3.4

LocalHero [2011−01−29 12:40:03/2011−01−29 12:55:57]

Last 3.3Bollinger Bands (20,2) [Upper/Lower]: 3.383/3.262

overround :0.014

0.000

0.005

0.010

0.015

0.020

0.025

0.030

Jan 2912:40:03

Jan 2912:42:00

Jan 2912:44:00

Jan 2912:46:00

Jan 2912:48:00

Jan 2912:50:00

Jan 2912:52:00

Jan 2912:54:00

Jan 2912:55:57


Market Probabilities to Win Prior to Race

plotProbs(bp1)

0.0

0.2

0.4

0.6

0.8

1.0

Time

Mar

ket p

roba

bilit

ies

Mark Twain Third Intention Lapin Garou Maoi Chinn Tire Akula Local Hero Indian Daudaie

12:40:00 12:42:00 12:44:00 12:46:00 12:48:00 12:50:00 12:52:00 12:54:00


Market Fundamentals Data

The Betfair API getSilks function provides basic ancillary datafor horse racing including age, breed, weight, days since last run,trainer, and jockey.

Colin and I also provide cheap subscription access to years ofdetailed historic market and fundamentals horse racing data,updated daily at:



An Example R-based Automated Trading Platform

Betfair

Market data collection script

Data StorageModeling and signal generation script(s)

Betting and account management scripts

• Three sets of independent R scripts.• Signals published via new Redis Publish/Subscribe or

Websockets.• Betting scripts run locally by users.


Lots of Opportunity for Quantitative Methods

Gambling and mathematics have a long mutual history. The abilityto analyze and act on high-frequency real time betting markets isrelatively new and pretty interesting.

• Bias detection and classification (favorite/longshot bias).• Arbitrage detection.• Handicapping models.• Estimating optimal bet allocation.• Backtesting and simulation.• Time series analysis.


On Deck

• This will be on CRAN and http://www.betwise.co.uk soon.• Games exchange REST API (a pokeR bot perhaps?).

USA residents: Please, write your Congressman and demandlegalization of Internet gambling in the United States.


Documents

The betfair package: An R implementation of the Betfair APIpast.rinfinance.com/agenda/2011/BryanLewis.pdf · The betfair package: An R implementation of the Betfair API Bryan Lewis