Now Available! EBook from CWM

Now Available! EBook from CWM
Order Now from Amazon

You can also get an epub copy


Monday, November 17, 2008

Statistics for Womens’ Curling and What is “Control”?

We’ve finally gathered data for Women’s Curling events. This data is taken from 4-rock games played during Provincial, Scotties (i.e. Canadian National Championship), World Championships, Olympic Trials, Olympics and WCT events over the last several years. Unfortunately our sample size is larger (more than double) for the men’s events. However, we have enough numbers to give some indication of general trends and comparisons with the games of their opposite sex counterparts. So what do we find?

Tied with Hammer

Within 1-2% of the men until the End Game (last 3 ends). During final ends, winning chances for women are 60, 61 and then 69% in the final end vs. 65, 66 and 75% for men.

1 Down with Hammer
Of all the scenarios, this is most similar to men’s numbers. Usually only 1-2% better chance for women to overcome the deficit of 1 down with hammer to win. Final three ends have same pattern: women’s is 39% to 37% to 39% (38, 35 and then 38% for men).

2 Down with Hammer
Women’s teams are again 1-2% more likely to win in this position except for the final end where data shows a 14.5% chance for a women’s team versus 10% chance for a men’s.

3 Down with Hammer
A 2-4% better chance for women’s team throughout the game.

Down 4 or more with hammer
Shows generally 2-4% increase in chance for women’s team.

1 Up with Hammer
A 2-3% better chance of holding this lead for men’s teams than for women’s.

2 Up with Hammer
Within 2-3 % early on but actually men’s teams hold on to win 4-6% more often from the Middle Game onward, with the exception of the final end where the difference is only 2%

3 Up or more with Hammer
Within 2-3 % early on but actually men’s teams hold on to win 4-6% more often during the Middle Game. The results from the End Game (final 3 ends) are nearly the same – you win nearly every time if you are in this spot.

What does this tell us about Women’s Curling?

Women’s teams tend to have a higher chance of coming back from a deficit and, subsequently, less chance of holding onto a lead. The difference is less noticeable in close games (tied or 1 up) and tends to widen as one team takes a more dominant position. That is, the greater the lead the greater the difference versus men’s teams in likelihood of a comeback. If we believe the ability to throw heavy peel weight successfully is the major difference in women’s and men’s games, then these numbers look much like what we would expect.

The most noticeable and important difference seen is the case of tied with hammer or down two with hammer in the final ends. The difference is about 5%. These numbers provide some support to determine how women’s teams might approach the game differently than men’s teams. For example, if tied with hammer with 3 or 2 ends remaining, a women’s team may be less inclined in blanking to retain hammer than in forcing a score. Tied with hammer is, during these ends, not a statistical advantage over 1 up without. In fact, with 2 ends remaining, women’s teams have a slight (2%) statistical edge in being 1 up without hammer! In the men’s game tied with hammer is an advantage of 3-4% with 3 ends remaining and 1% with 2 remaining.

What is Control?
A common term heard over drinks at curling rinks across the globe is “Control”. “If we make this shot for two that will put us in control”. “You had control from the 5th end on”. And, the most hated phrase “We had control the whole game and lost it at the end”. So, statistically speaking, what do we think is meant by “Control”?

I propose that there are actually three positions during a game a team can be in. If one is Control it follows that there must be another type of game that is closer than this, which I will call a “Close” game. It also then is reasonable to suggest there is a position where you are even better than in control, let’s call this “Dominant” position.

If we assign the game condition based on a probable outcome:

Close occurs when the odds for a win is no greater than 66% for either team. Another way to say this is the team behind has better than 2-1 odds of winning.

Control exists when one team has greater than 66% but less than 80% odds of winning. This range is between 2-1 odds and 4-1 odds for the team that is behind.

Dominant position is when one team holds a greater than 80% statistical chance of winning. Another way to show this is greater than 4-1 odds of a comeback.

Based on these numbers:
Control” occurs when team is:
Tied with hammer in the final two ends or extra end.
Up 2 without hammer anytime before two ends remain.
Up 1 with hammer anytime before the final four ends

With three and four ends remaining, 2 up without hammer is right at 79 and 80% respectively. With two or fewer ends left, you are Dominant in this position. When statistically in “Control” Up 1 with hammer, you are between 77 to 80%, with the exception of the third end in a ten end game where your chances are 75%.

Close” position occurs when a team is:
Down 1 with hammer
Tied with hammer anytime before the final two ends
When tied with hammer and 2 ends remaining, chance of winning is exactly 66% - so we could argue whether a team has Control at this stage or it is Close.

Dominant” position occurs at any other score during the game.
We can start to analyze our pre-game and in-game strategy using the definitions of Control, Close and Dominant. I’d also suggest incorporating the definitions for which section of the game, based on Early, Middle and End Game (from article of Nov 2008, “Is Curling a Battle For Hammer?”). Recall the End Game is the final 3 ends and extra end if required. The Middle Game is the middle 3 ends and the Early game is then either 4 ends for a 10 end game or 2 ends for an 8 end match.

Using this model creates the following 9 game scenarios:

I’d suggest teams could use this model as a way to develop pre-game strategies for how to approach each of these positions. I might even tack a stab at examining these, but that will have to be left for another article…

Statistics for Grand Slams
Watching the Masters a short while ago (except the semi’s which were on the BOLD network –wherever that is), I started to question the statistical basis we are using and see if there are some discrepancies for the Slam events versus the entire dataset for all WCT (including Slams), Olympics, Olympic Trials, Brier, Worlds and Provincials. Our data size is still very small but I wanted to get a general sense if we saw any differences. Tied with hammer in final or extra end is currently nearly 80% for Slams versus 75%. 1 down with hammer in 9 was 39%, close to our baseline of 38%. Down two with hammer was about 12% versus 10%. The only major difference appears to be tied with hammer. The Slams numbers are based on sample size of about 270, compared to over 3000 in our full dataset. Assuming that Grand Slam teams are generally stronger than the other fields; does this mean better teams win more than 75% when tied with hammer? Possibly yes, but difficult to say for certain without a larger sample.

If Martin played Howard a single end game 20,000 times, each team having hammer for 10,000, what do you think the percentages would look like?

Sunday, October 26, 2008

Is Curling a Battle for the Hammer?

Back during the era of the Grand Old Game; before push brooms, The Ryan Express and Shorty Jenkins ice moved us all to Free-Guard Zone; it was often heard at most every club by players at every level (of play and inebriation) that “Curling is a battle for the hammer”. To some degree this may have been true, though the strategies of teams like Savage and Werenich, Burtnyk and Olson and the Howard brothers opposed this mentality with their aggressive play. Rather than “Take 2 then Give 1”, these Superteams preferred to put constant pressure, take many and then steal until the opponent shook hands.

The question I raise is, in Today’s 4-rock game, is it still possible that Curling can be a Battle for Hammer? In my analysis, I will touch on several areas and provide some data which may help us find and answer.

The Hammer Advantage
Starting with hammer is an advantage. We know this because every time your team wins the toss, you don’t think too long about whether or not to choose stone colour. I expect some Ontario folks will shout out “I remember this one time so and so knew the rocks were bad and chose colour instead of hammer” – but I expect this is not a common occurrence.

The “draw to middle” or pre-selecting number of games in a round-robin where teams get hammer also indicates its importance.

We also have some data which tells us how often the team beginning with hammer wins. Unfortunately not all events keep proper track of who started the game with hammer and our sample size is not as large as we would like. One way to take a sample of data is looking at 10 end games that are tied after 2 ends (i.e. now an 8 end game). Our results show us a 61% winning percentage.

This doesn’t tell us much more than we already know, except to help us understand if 8-end games are fair. That is not, however, the purpose of this article (though we may return to this idea).

The Early Game
I believe Curling, sometimes called “Chess on Ice”, has an Early, Middle and an End Game (no pun intended, honest). End Game is the final 3 ends and extra end if required. The Middle Game is the middle 3 ends and the Early game is then either 4 ends for a 10 end game or 2 ends for an 8 end match. I haven’t chosen these game sections based on my best guess or what I’d like them to be, I’ve looked at the numbers and they reveal something very interesting:

The Odds of winning at the completion of each end during the Early game is nearly equal.

This is a fascinating discovery that not only explains why the Early Game doesn’t end until 6 ends remain in the game, but supports the theory that an 8 end game is competitively equal to a 10 end game. If you extrapolate the numbers, we might be even safe in assuming 12 ends or even 14 would also have the same outcome. The only benefit of a longer game (other than perhaps more beverage sales to the fans) is that the more ends played, the greater advantage a stronger team will have over a weaker team. The analysis behind that theory is not the purpose of this discussion, so back to where we were…

Middle Game
Odds of winning during the Middle Game start to trend in a specific direction.

End Game
Odds of winning in during the End Game trend steeply in one direction. The exception is 1 down with hammer, which drops then rises before the final end.

This graph best shows this trend:

Some comments:

Tied with hammer: Winning percentage is within 60-62% until 3 ends remain where it jumps more dramatically to 65%. That is, a tie game has nearly the same statistical outcome for all ends until the End Game is reached (starting the 6th or 8th).

1 down with hammer: Maintained around 42-43% until 5 ends remain, where drops to 39% in the Middle Game. It then rises back over 40% and during the End Game, 1 down winning percentage drops to 38%, then down below 35% and back above 38%. This phenomenon was discussed in the article “To go for two… or not? Masters of Curling final: Howard vs. Ferbey” from Black Book of Curling 2007-08.

Two down with hammer shows 27 to 26 % during Early Game than a drop below 25% beginning the Middle Game and a drastic drop to 20% with 4 ends remaining.

3 down with hammer: Stays around 15% until 6 ends remain then drop to 13%, then flat at 11 to 10% with another drastic drop beginning the End Game of below 7%.

1 up with hammer: Stays fairly constant, from 75% then leveling in the range of 77-80% until the End Game begins and it jumps to 84%.

The other scenarios are not as common, but do have some interesting results:

4 down with hammer: Transitions, oddly enough: 9-7-7-5-4-3-2-1-0%. Short answer – don’t be 4 down after the Early Game or you’re screwed.

2 up with hammer: 87-90% until we reach the final stage of the Middle Game (93%).

3 up with hammer: In an 8-end game, anytime you are in this position you shouldn’t lose. In a 10 end game – just get past the 3rd and it would take a monumental collapse to lose (though we’ve all been there once or twice).

Battle For Hammer?

To try to answer our original question, we need to define what is meant by “Battle for the Hammer”.

Let’s assume “Battle for Hammer” implies a team which starts with hammer wishes to keep it and the opposition is trying to gain that position (tied with hammer). The advantage of tied with hammer is only slight, roughly 60 to 40, until the End Game. Teams who win 60% of their games don’t often place high in the money and they certainly don’t win Briers or Olympic Gold. We don’t gain a substantial position until the final end, where we still lose 1 of every 4 games. By this definition, I’d suggest Curling is not a Battle for Hammer.

Now, if assume we this phrase to mean a battle to gain hammer with the lead, then it could be argued Curling is a Battle for Hammer. 1 up with hammer with 8 ends remaining is the same as tied with hammer at the end of the game (75%)! Having hammer with a lead of 2 or more points is very strong. It is preferred to 3 up without during every stage of a game, except for the final end where both positions are equal.

How Important is the Hammer?
So how do we begin to analyze the importance of last rock? At any time during a game we can determine its statistical value. We’ve determined that leading with hammer is a strong position, stronger than being further ahead without hammer. But by how much? Let’s compare tied and 1 up without, 1 up with and 2 up without, and 2 up with versus 3 up without.

Tied with hammer vs. 1 up without hammer
Often, in a tie game, when a team is forced to 1 we state the opposition has done their job and taken away control. However, stats show us that there is only a small difference between tied or 1 up without
During the Early Game the difference is about 2-3%. It jumps to 5% at the beginning of the 5th end, or Middle Game. Then ranges between 0-3% until the last end where it jumps to 12% advantage for tied with hammer.

1 up with hammer vs. 2 up without hammer
When 1 up with hammer you are stronger than 2 up without by 2-4% until 4 ends remain, where it reverses to 1% advantage when 2 up, then back to advantage of 4% to 2% then 0% for the final end.

2 up with hammer vs. 3 up without hammer
Again, up with hammer is slight advantage, usually only 1% with the exception of 7 or 4 ends remaining where it is 3-4%. In real terms, these two positions are essentially equal.

So, what does this mean?
There is clearly not a significant difference in each of these scenarios. In each case having hammer while up is a slight advantage, but usually only 2-4%. Therefore, I disagree with the theory that Curling is a Battle for Hammer. Take, for example, an 8-end game where you have hammer and are held to one in the first end. Instead of having a 60% winning percentage you drop to 58%. Your position is in fact not much different than where you were at the beginning of the game. Much more significant is to have a shot for one and instead give up a steal in the first end, going from 58% to 42%.

Interesting to point out that often when the team without hammer holds the opposition to one in the first end it is perceived they have “won” the end or done their job. In reality, they have only gained a 2% advantage from where they were. More correct perhaps to state they have successfully “avoided” the position of falling behind by two or more.

Next article I will be revealing data on the Women’s Game and also attempt to tackle the question of what is “Control”.

Thursday, January 17, 2008

What strategy should I employ when one down with hammer in 9th end?

One of the most difficult situations towards the end of a game is one down with hammer playing the next to last end (9th or 7th). Statistically, the lowest probability of winning when one down with hammer is in the next to last end (34.9%). In the last end or with two ends remaining it is over 38%. In fact, tied with hammer with two ends to play is 67.5%, only 2.4% higher than if you are one up without!

I have seen every type of play in this situation, from both teams keeping it clean to produce a blank (and a 5 minute end) to every rock in play. So, using mathematics, what is the correct strategy, or at the very least, how do we approach this scenario to be better prepared when it happens?

Recall statistical outcomes for the final end:

Expected Results (ER) in the final end:
Odds of winning if tied with hammer (x) = 74.5%
Odds of winning if one down with hammer (y) = 38.2%
Odds of winning if two down with hammer (z) = 11.0%

So how should we play the end to maximize our chances and overcome our unenviable position? Also, our opponent without hammer can dictate the early part of an end by placing guards, how does this impact our decisions?

Let’s start by examining the numbers. Clearly, the best scenario is to take three (or more). If we score three we have an 89% chance to win. The next best scenario is to score a deuce and have a 61.8% chance in the final end. If we play aggressive and are forced to one, however, we win only 25%. In fact, if the end results in a draw for two and we miss, only scoring one, our chances flip from 61.8% to 25% - a huge difference. A blank is better than scoring a single, leaving us at 38.2%, but still we can expect to lose more than half the time.

Let’s also examine the competition’s strategy. If they read my articles, they know the correct play in this situation is to force the action and attempt a steal or force or a single, at the risk of a deuce.

Let’s assume for now our opponent will place a centre guard.

Option 1:
First, let’s attempt to play a clean end with the expected outcome a blank. After the opposition places a centre guard, we choose to draw to the side. Our opponent will most likely hit our rock and stay in the rings. Assuming we exchange shots and no one rolls their shooter to center, we can remove the centre guard with the 6th rock of the end. If the opposition splits the rings (most likely) we then are playing out the end trying to make a double in order to blank. If we fail to make a double and our opponent does not roll out, we will be forced to a single point and left with a 25% chance of winning with one end to play.

This option appears to be a losing strategy. No chance for three and not likely two, so our only outcomes are less than 50%. Let’s estimate some outcomes based on this strategy:

Score Three = 0
Score Two = 5%
Score One = 30%
Blank = 60%
Steal = 5%

W = 34%

Option 2:
Let’s try a very aggressive strategy. Come around with our first rock, corner freeze if our opponent does also, and continue drawing or soft taps until an opportunity develops for a “big shot” at scoring multiple points. Let’s again make some rough estimates based on this strategy:

Score Three = 10%
Score Two = 30%
Score One = 40%
Blank = 0%
Steal = 20%

W = 40%

Even if we could successfully blank 100% of the time, we do better playing very aggressive.

These numbers are not completely fabricated; they are based on existing data. For example, as of this writing, during this season (2007/08); Howard, Ferbey and Martin score three 13% of the time across all ends played with hammer. Taking an average of three “average” WCT teams results in 8%. These numbers can be skewed due to the better teams (aka Howard, Ferbey and Martin) being ahead more often and their opponents need to take greater risks – often resulting in big ends during later stages in a game.

Option 3:
Let’s think of a third scenario. We come around the guard, buried in the top eight foot, possibly biting the four foot. Our opponent successfully corner freezes. We now have several options:

Corner freeze to the opponent stone
Draw to the side of the rings
Split the centre guard with a “tick”.

Shot Call 1 will lead us to the scenario in Option 2 above. What about the other two?

2. Draw to side of rings:
If we can sit second stone, our opponent has several options, depending on how all the stones are sitting. In most cases, we could expect he will attempt a hit on the open rock and try to roll to the centre, behind cover, to sit either first or second. If he is successful, we are back to Option 2 (Very Aggressive) above and we are likely behind in the end. We will need a big shot or mistake from our opponent, but the aggressive nature of the end now makes that more likely. If our opponent does not roll successfully, we can attempt a hit and roll. In either case, if we instead choose to remove the guard or run it back, a steal becomes less likely, however a chance for a single increases and a three is highly unlikely. We can expect the end will look more like Option 1 above, with our best outcome a blank or low probability of a deuce.

3. Split the centre guard:
This is the scenario which I don’t recall seeing in a game but appears powerful. If we can split the guard and create two corner guards, our opponent now is left with an unclear decision. Does he put a centre guard back, even though you’re shot stone? Does he peel a single corner, or attempt a double peel (if it is even possible). Does he run his stone onto yours, attempting to lie two? This last option seems the best scenario but, it will leave two guards, an open four foot, and most likely both rocks will not sit perfectly behind cover above the tee line. Even if they do, a corner-freeze is available and with two guards, one which is now yours and could be driven back later on, the advantage appears to be with you. If our opponent gets cautious and elects to peel the guards, we now have options to play for a blank (if we choose), place another guard or move rocks around in the house and attempt our deuce with minimal chance for a steal.

Ultimately, a team must determine which option maximizes a chance for two or three, limiting the times you get a single point AND minimizes a steal by your opponent. Not an easy answer. The difficulty is that a blank, which is preferred to a single, is not a high occurrence if you attempt to score two or three. This specific scenario, one down with hammer with two ends to go, is one of the most interesting in the game and possibly more intriguing than the final end.

What If Our Opponent draws into the rings?
Instead of a centre guard, our opponent puts his first rock in the rings. We can now place a corner guard or hit the rock in the rings. If we call for a guard, our opponent could now choose a centre guard. We then draw around and are back to scenario in Option 2 above, though we are behind in the end. If we hit the stone, we are playing for a blank which, in this case, is very likely. For the team without hammer, assuming we will place a corner guard, this appears to be a stronger play than above. The team with hammer now faces the corner freeze and could have more difficulty getting shot stone.

This decision stems more from a team’s strategy of preferring to sit shot or be positioned frozen to shot stone in order that a shot can be played later in the end. I can see advantages for both cases and will leave it up to the reader to determine what they prefer in this situation.

Appears there is no clear answer to our original question. It is clear that attempting a blank is a less risky play but provides no upside and most likely results in us winning less than one in three times. Playing aggressive increases our chances, but also creates a complex end; producing many options for both teams and presenting opportunities to stay aggressive or bail out. For fans, it is clearly the most interesting scenario in a game and, for a skip, one that cannot be simply “played by the book”.