Cortana made me a better gambler, but I'm never listening to her again
I'm a bit of a gambler. I've probably placed at least one bet per week for the past year. I'm also pretty good at it -- since winning a few hundred dollars on the World Cup last year, I've maintained a positive account balance at Ladbrokes (other betting sites are available), never adding additional funds, and never betting more than $10 on a single event. Last week, Cortana began predicting English Premier League results, so I decided to see what would happen if I let the Windows Phone (and soon to be Windows 10) assistant make my decisions for me.
I bet £5 ($8) per game on a total of 14 games from the Premier League. Those games were all the 3PM kickoffs on Saturday February 7th and all the evening games on Tuesday 10th and Wednesday 11th bar one. And I'm actually up £24 ($37) -- but as I'll explain, it's probably not time to get a Windows Phone and quit your day job.
Ten out of the 14 bets I asked Cortana to call were correct. I lost £20 ($31) from four bad bets, and won £114.42 ($175) from the 10 good bets, leaving my balance at £94.42 ($144), or £24.42 ($37) higher than what I gambled. That's a 35 percent return on my investment.
Here's my balance sheet for the week:
My week of gambling with Cortana | |||||
---|---|---|---|---|---|
Man. City vs. Hull City | Man. City | Nope | £5 loss | £5 | -£5.00 |
Swansea City vs. Sunderland | Swansea City | Nope | £5 loss | £10 | -£10.00 |
Leicester City vs. Crystal Palace | Crystal Palace | Yes | £11 win | £15 | +£1.00 |
Aston Villa vs. Chelsea | Chelsea | Yes | £2 win | £20 | +£3.00 |
QPR vs. Southampton | Southampton | Yes | £4.50 win | £25 | +£7.50 |
Arsenal vs. Leicester | Arsenal | Yes | £1.43 win | £30 | +£8.93 |
Hull vs. Aston Villa | Hull | Yes | £6.50 win | £35 | +£15.43 |
Sunderland vs. QPR | Sunderland | Nope | £5 loss | £40 | +£10.43 |
Liverpool vs. Tottenham | Liverpool | Yes | £4.75 win | £45 | +£15.18 |
Chelsea vs. Everton | Chelsea | Yes | £1.82 win | £50 | +£17.00 |
Man. Utd. vs. Burnley | Man. Utd. | Yes | £1.67 win | £55 | +£18.67 |
Southampton vs. West Ham | Southampton | Nope | £5 loss | £60 | +£13.67 |
Stoke vs. Man. City | Man. City | Yes | £3.50 win | £65 | +£17.17 |
Crystal Palace vs. Newcastle | N/A | N/A | N/A | £65 | +£17.17 |
West Brom. vs. Swansea | West Brom. | Yes | £7.25 win | £70 | +£24.42 |
Now, the huge caveat: I only bet on five out of 10 games over the weekend. I didn't want to put more of my own money into the black hole that is gambling, so I went with just the 3PM kickoff games. As it turns out, the five games I didn't bet on all went against Cortana's predictions, and because of that I decided to bet on all 10 midweek games to try and get a clearer picture.
I took the break between games to examine exactly how Cortana is working out these calls. She pulls her predictions from Microsoft's Bing Predicts search engine feature, which has been live since April last year. It initially called reality shows like The Voice before moving onto the World Cup, the NFL and finally the Premier League. For sporting events, it looks at what Microsoft calls the "historical strength of teams." That means the number of wins and losses, margin of victory, location of matches, weather conditions and more. It also adds in "the wisdom of the crowds," by pulling in data from social media and other "web activity."
By searching for predictions on Bing rather than Cortana, you actually get a little more information on the supposed likelihood of each prediction coming true. Asking Cortana for last night's results, she firmly called victories for Manchester United, Southampton, West Bromwich Albion, Chelsea and Manchester City. On Bing, you get percentage figures rather than outright predictions. There was a 71 percent chance United would win, for example, and only a 35 percent chance West Brom would win. Cortana doesn't offer this level of detail, however, and only predicts winners and losers. It's because of this that I was unable to place a bet on the Newcastle United vs. Crystal Palace game. Bing called a draw for this game -- which actually happened -- but Cortana flat-out refused to answer my many requests for a prediction. It would be great if Microsoft could add the "draw" call (which seems to come very rarely) to Cortana's playbook. If she had this functionality, I would've actually won more money.
Bing correctly called a draw for Crystal Palace vs. Newcastle, but Cortana refused to give me a prediction.
If I had bet on everything -- the five games I didn't bet on over the weekend, and the Crystal Palace draw that Bing predicted -- I would've ended the week with about $1 profit. If I had ignored Bing's call and only gone with Cortana, I would've been around $1 down. It's random chance that I initially chose the 3PM games to test this out with, and ended my week of degenerate gambling $37 up. It's pretty apt, wouldn't you agree? Despite Cortana and Bing's advanced data analysis, whether I won or lost came down to what all gambling boils down to: luck. And I'd rather make my own.