What are the odds?
Imagine you're a bookmaker, or better yet, a trader. The day is May 15 2020, the game is Counter Strike: Global Offensive. G2 Esports are facing Team Vitality
in the lower bracket final of the prestigious ESL One: Road to Rio EU event.
Coming into the match, G2 was the higher ranked team (ranked 2 in the world), but Vitality (ranked 10) had shown a better performance. Experts were predicting
Vitality to win, bookmakers were split - most were offering 1.8/1.9 odds
in prematch, with either team being favoured equally often. One thing everybody seemed to agree on, this match could go either way.
Map 1 was Nuke
, and picked by Vitality. G2 eked out a close win with 9:6 at half-time and 16:14 overall.
Now we are on map 2, Vertigo
, and G2 is leading 10:3. They are playing terrorists, bomb has just been planted. Which odds would you offer on G2 to win the map and match? At this point, it seems like a done deal. Offering 1.1/10 does not appear unreasonable.
Well, if you've done this, you might have lost a lot of money. Vitality proceeded to win the remaining two rounds in the first half, and then took the second half, 11:4, making the overall 16:14 in rounds and 1:1 in maps.
This means we will see a third map. Dust2 is the decider.
Who do you think will win now? Do you expect Vitality to continue their momentum? It sure looks like that, because now it's Vitality who develop a 10:3 lead! Surely the game is finally almost over? But no, G2 come back, tie, and finally win 16:14. It's 2-1 for G2, who advance in the tournament.
What a crazy and exciting game! What a delight for spectators! But how did the betting industry fare? Without disclosing privileged information, we can safely say that there are bookmakers who lost and won on this match. A lot. Enough for us to take a closer look into what might have happened and why.
Comebacks in numbers
Both Vitality and G2 each came back from a disadvantage of 7 rounds. It's easy to file this match under "extreme examples of bad luck" and move on. But was it really so extraordinary? Let's look at the statistics of CS:GO. In our data set we have around 15,000 matches that happened after the March 2019 economy patch. These are all professional CS:GO matches, ranging over various skill tiers and regions. Let's look at the largest lead a team had in each match, and whether or not their opponent was able to come back from it. We have plotted the numbers below:
We see that an advantage of 5-7 rounds is the most common in CS:GO. The comeback percentage goes down dramatically as the lead in rounds increases. The seven round advantage that we saw in Vitality vs. G2 happens on 13% of the maps (a lead of at least seven rounds happens on 60% of the maps). Statistically, the losing team came back in only 14% of the matches. Assuming a moderate take of 0.08, that's odds of 1.25/7.7. Not the 1.1/10 we had suggested at the start of this article, but not exactly high, either. If this is the only statistic you look at, you will say that it's quite unlikely for a team to come back once they are seven rounds behind.
Half-time as a clean slate
We'd like to argue that even these odds are too low, statistically speaking. What we have not considered so far is when the seven rounds lead had happened. Since the score was 10:3, we were still in the first half. And here is the thing: the statistical probability of a team coming back from a whopping seven round disadvantage is almost 25% if the score is reached in the first half!
This make sense considering that CS:GO is played in two halves a 15 rounds. At the start of the second half the economy is reset and a pistol round is played a gain. If you ignore the psychological pressure of being behind, it's a clean slate for both teams. In the Vitality vs. G2 match, on both maps the score was 10:3 when the turning point was reached - the losing team could go all out in the first half, securing the remaining two rounds. They then played brilliantly in the second half, winning the match.
We can take this idea even further by asking ourselves what a large lead can mean. There are two distinct scenarios. Scenario 1 has one team dominate over the other. They are expected to win with a high margin and it's likely that their opponent will performing as badly in the second half as in the first half. The maximal lead in rounds is achieved in the second half and there is no coming back from it. Extreme odds are justified.
Compare that to a situation where one team was just unlucky - maybe they lost a couple of rounds in a row, had a rocky start, but there is every chance of them coming back. Without going into details of how CS:GO economy works, losing a couple of rounds in a row might even constitute a tactic. This is something we expect when two teams are rather evenly matched. In this case we still want to attach a higher win probability to the winning team, but it should not be as extreme.
Do sides matter?
In CS:GO, one team will start out playing terrorists and the other team will be counter-terrorists. At half-time, they switch. Fans will argue to the death which map favors which side - even in Bayes team we are not in agreement about this, though statistics show a small CT advantage for both Dust2 and Vertigo.
This advantage becomes a lot more prominent when we look at comebacks. Both G2 and Vitality had started out playing terrorists. By half-time they were five rounds behind and played CT. For both maps, starting with a 5 round disadvantage, the comeback probability for CT roughly 25%! That's significantly higher than for terrorists, where it sits at around 15% and A LOT higher than the 14% we had our analysis with - if we follow statistics, we are now offering something in the realm of 1.4/4.3.
Staying on the safe side
Vitality vs. G2 was exciting not only because of the two comebacks, but because there are bookmakers who made or lost a lot of money on it. There is no magical trick to avoid losses - and sometimes it is just one lucky bet that can offset your entire book. The message here is not "your model is bad if you lost on this match" - a lot of bookies did. It's "don't make the odds too extreme". What looks like a clear-cut win for one team is less likely that it subjectively appears. Knowing the teams and being an active fan of the sport also helps, of course. In this case, two aggressive, evenly matched teams with proven nerves of steel were facing off. If we make a list of teams that historically came back from a large disadvantage, both Vitality and G2 make the top 20. That by itself made a series of comebacks more likely.
This match stands as a statistical outlier in terms of betting returns - and that might be a sign that the betting industry is still learning to price esports matches correctly.
Unfortunately we do not have map winner odds for this game that we can publicly share. Watching the VODs, however, we can see odds of one bookmaker offered at half-time. These are (presumably) match winner odds, but we can still use them as an example. We are not saying whether or not they were successful with their odds, we are only considering the numbers. As a bookmaker primarily focused on esports we expect them to have a good idea about the game.
In prematch, the bookmaker was slightly favoring Vitality at 1.9/1.8. This prediction shifted quickly when G2 developed a lead. At the start of the second half of Nuke, G2 was leading 9:3 and the bookie was offering odds 1.76/2.01
, which is still rather balanced. Fast forward to the second half of Vertigo: G2 is leading 1:0, 10:5. The bookie is offering 1.14/5.32 for G2
to win. That's a 0.18 probability for Vitality to come back on this map and then win the next map, meaning the underlying probability for Vitality to win Vertigo is somewhere in the 0.25-0.35 range. That's even higher than our analysis suggested! A Vitality fan would happily take the match winner bet - and make them some money.
In Map 3, odds are shown again at half-time. Interestingly, they are now 2.57/1.48
- 0.36 probability for G2 to come back! Since we are on the decider, map and match winner are the same. Here is when in-depth knowledge of the two competing teams comes into play - the probability of G2 coming back is simply higher than for the average team. And ultimately G2 did win, so this bookie had it right.
Where they were wrong was to favor Vitality in prematch. This is another way for a bookie to lose a lot of money - overvalue Vitality in prematch and have all G2 fans (and there are loads) take you up on that.
Ultimately, statistics can only get you that far. It is the quality of the matchup happening, the skill of the two teams playing and sometimes that pure luck that decide the outcome of a match and who you should bet on.
About the author
Dr. Darina Goldin is the Director Data Science at Bayes Esports. She started playing competitive Team Fortress 2 in grad school. While no longer competing, she is still an avid Esports fan. At Bayes, she has created numerous predictive models for Counter Strike, DotA2, and League of Legends. When not crunching numbers, you can find her at the gym training Brazillian Jiu Jitsu.