We’re going to talk about the tournaments and why going anti-meta might be the best route to victory.
This post draws on some of the points discussed in previous parts of this series. I’ll link backwards to more thorough exploration of the concept at the game-level, but this should stand alone.
But first: It’s not all about winning.
Before we start talking about winning, I think there is a much more important point.
Most of us won’t win the tournament, but we all go to have fun.
This is the trophy I’m most proud of:
It’s the Kilted Kiwi award, given to the player voted as “most sporting” in terms of fun/spirit of the game.
In that particular event I went 2-4-0, finished mid table, but went home happier than ever, knowing that my opponents had enjoyed their days as much as me. Win, lose or draw this is a good target that we can all play toward at every tournament.
Learning to enjoy yourself and make it enjoyable for your opponent irrespective of the outcome is more important than any trophy or ranking points.
Throughout this post I’ve asked for input from successful coaches – see comment at the end from BB_Jock that speaks more to this point.
Optimising for Expected Value or Outcome Distribution
Expected Value is the term for the average outcome of a decision across hundreds or thousands of repetitions. A point we’ve made early and often in this series is a route to improving is aiming to maximise expected value accross actions, sequences, drives and whole matches. Making optimal decisions as often as possible and you will more games. On average.
The outcome distribution is the shape of the curve; does it have long tails or is it a narrow peak. The width and shape of that curve is defined by the variance profile, how much of the outcome is dependent on chance.
The outcome distribution is therefore shaped by your playstyle (if you cage dive every turn it will be a wider curve than if you wait for chances), and by the race/roster (a glass cannon team might have a wide distribution on whether they bang or get banged, whereas a draw-heavy Nurgle team might have a tighter distribution with fewer wins and fewer losses).
Over enough games, stronger rosters and stronger approaches to the game and stronger player tend to generate better results.
None of that is a shock.
However, “Over enough games” does a lot of heavy lifting in that sentence…
Tournaments are not played over hundreds of games; they are played over three, five, or six or occasionally nine.
A one-day Tournament is a high variance event
In a one day tournament the route to victory of the event can be very different.
Round 1 could pit you against someone playing their first ever tournament or against the England Captain (Thulean, great player). Win that somehow, and you could face the Scotland Captain (BBJock, also, great player), or you could face the winner of the game between the two fun-having Goblin coaches who were paired together in game 1, and then in game 3 you could play AndyDavo (current world #1 by Elo) on table 1 or someone else who came by a similarly favourable route on table 2*.
*The CLAW boys take great pleasure in pointing out that my first tournament win involved playing vs Snotlings, then more Snotlings, and then avoiding Kfoged in the last round. Don’t knock it, it worked!
Whoever you face along the way, we all know anyone can win any one game: you get hot dice, kill 4 on turn 1, land your TTM one-turn etc.
The longer the format, the more variance stabilises and true EV matters. By round 4 of a 2-day the top of the field will mostly be made up with the consistently strong performers, and the eventual winner will likely be someone you know. By game 6, just being lucky isn’t usually enough to put you on table 1, you do need to be a strong player too.
On that note, 2-day tournaments are much better for gauging your skill level, progress as a coach, and to get more valuable data on whether an off-meta pick is viable.
Tournament victory is a “tail event“
Pretty obviously, most people who attend a tournament don’t win it. And, for the most part, wins are shared around.
I’m writing this in Sept 2026 and so far this year in UK tournaments, AndyDavo* has won 6, Jairo has won 5, three players have won 3 (Scouseboy**, Chuck Dredd and Wacky Willow), 18 have won two tournaments, and 96 different coaches have won one.***
*I got Andy to comment on this article, his quote is further down the page – jump to it via this link
**I also got Craig, Scouseboy, to comment. His quote is toward the end of the article – here
***I’m in the 96 (1 win, 1 runner up, 1 team tournament win). Part of the motivation behind writing this post comes from wondering whether I should change something to win more.
The graph above attempts to capture this conceptually. The people to the right of the box are the ones who are in the 2, 3, 5 category or will end up there by the end of the year. The ones in the box are the rest of the one-winners (+ the same number again that have finished top 5).
Relative to the skill of the rest of the field, if you think you sit in the boxed region then yes you will win tournaments sometimes but you already know you need a good day. The right matchups, reasonable dice and to play well.
Playing to optimise EV in every game will push your average up. You will finish more tournaments in the top half, top quarter, top 10% as you become better at optimising EV. But, in terms of winning tournaments, it is the distribution of outcomes that matters more than the average outcome.
We’re going to come back to this shortly, but now a quick aside.
Different Coaches Are Solving Different Problems
The mechanism to maximising your tournament win chance depends on your relative skill level compared with the rest of the field.
A top coach derives much of their advantage from superior decision making during the game. They do not necessarily benefit from injecting variance, unusual matchups, or surprise factor. They benefit more from reliable rosters that allow their skill edge to express itself repeatedly.
Their outcome distribution is already right shifted, more of the curve is in the happy multi-win territory that yields tournament wins.
Top coaches are usually best served by selecting the most efficient, consistent strategy available. They can still look for an edge in their build, but they don’t necessarily need it. Over a large number of games, and against opponents of varying strength, low-variance approaches that reliably convert small advantages into wins will tend to produce the strongest results.
When elite coaches discuss roster selection, (e.g. if you watch an AndyDavo or Jimmy Fantastic roster youtube video) remember that they might be solving a different optimisation problem to the rest of us.
Those roster pics are correct, and once you know how to pilot them, they will give the maximum win chance in any individual matchup.
But
If you aren’t playing at elite level, simply copying the elite choices may not be optimal in terms of your chance to win the event.
Growing your tail = winning more tournaments
Do you want the highest average finish, or do you want the highest chance of finishing first?
The problem of winning more tournaments becomes one of changing the outcome distribution. What we want to do is have more of the outcome distribution within the “tournament winner” portion of our graph, even if it shifts OUR average (EV) to the left.
To do this, we might need to trade a small amount of consistency for the opportunity to gain an edge, especially against the opponents (races/rosters) most likely to stand between us and the trophy.
However, adopting an approach that will win more tournaments likely means also adopting an approach that loses more games on average.
There will be more times when you go home disappointed and your ranking will suffer.
Off-meta or Anti-meta
Just a quick distinction in terminology. These two terms get used somewhat interchangeably but they are different in important ways.
An off-meta strategy might be something new and innovative. A skill pick combination or roster build that hasn’t been tried before/often/locally or could be a different way of playing the same team.
These can be impactful. Your opponents may not have baked-in knowledge of how to respond. The new idea bump is why new races and the teams with different mechanics can often do better than their stat lines might suggest, especially against inexperienced coaches. Think back to your first time playing against Slann and Vampires… and look forward to that first game vs Squigs.
When you are innovating, you do pay an innovation cost. Not only are your opponents reacting to something new but if it is new for you too, then how to make the most of the new build might take some reps.
An anti-meta strategy is looking at the current meta game and identifying a weakness, then choosing a race/skill combination (that might look completely normal) but which is designed to defeat the meta.
Defeating the meta does not mean bringing a bad or mad roster. We need a roster capable of beating weaker opponents consistently. It means identifying strong options that attack common assumptions in the expected field that the elite players will be using.
Exploitation vs robustness at the metagame level
In Part VIII of the game theory series I talked behavioural game theory and playing your opponent. At the game level, we talked about using exploitative strategies to gain extra value from specific assumptions about how opponents will play.* In contrast, robust strategies performed well regardless of opponent behaviour but don’t give a specific edge.
*If you want to see my tendencies being exploited… AndyDavo recently used an exploitative read in a game vs myself – described in this puzzle.
Exploitation and robustness transfer to the wider meta-game very easily.
As tournament metas mature, coaches often converge towards a stable equilibrium (in Game Theory, a Nash Equilibrium) of strong roster choices. That equilibrium may be difficult to improve upon, but it can still be attacked.
A robust team (or roster composition) pick will be strong because it is robust. Think Tier 1 team filled with safe elite-skill picks.
If we use this year’s Eurobowl as an example, this is your High Elves with Dodge on Princes and Lions, a Leader Thrower, an extra Wrestle somewhere, and probably a stack of Strip Ball or Frenzy on a Lion. It’s got everything it needs to win in any matchup.
The exploitative team pick involves reading the meta and identifying a weakness. The challenge is whether a perceived weakness is real or merely imagined.
Again, if we use this year’s Eurobowl as an example, it looks like all 4 Elf teams are viable and certainly have been over-represented at EB ruleset tournaments I’ve attended. An exploitative team pick might be rosters that have tackle/mb. It might be that teams like Undead, which might not look all that strong at first, are actually a strong counter to the meta.
Tackle mighty is obvs good against almost anyone. A different meta read might be taking Diving Tackle (a “Senor Big Hat”, a Blodge-step Diving Tackle Elven Union Blitzer perhaps?), or it could be saying that Elves don’t usually have Sure Hands, so Strip Ball increases in potential value in those matchups.
But perhaps we can do better than going off feels.
Game Theory Terminology Explained
A Nash equilibrium is a game state where no player can improve their outcome by changing strategy alone, assuming all other players keep their current strategies. In metagame terms, it represents a stable set of choices where each player’s strategy is the best response to what everyone else is expected to do.
Dynamic vs Stabilising Metas
To exploit requires information and leads to an important consideration:
If you have poor information, robustness becomes more valuable. If you have excellent information, exploitation becomes more attractive.
Which leads to the question… do you actually have strong information about the meta? While also remembering that the meta isn’t ever truly static, because everyone is reacting to everyone else.
Example – Dynamic meta reads
I think the Tackle / Dodge interaction is useful here as an obvious pick/counter pick pair. We can all appreciate that Tackle is a wasted skill if your opponent has no Dodge. If you are considering taking it as a chosen skill on your roster you are making a value decision on how much Dodge you will run into and how much difference having Tackle will make on your win conditions when you do.
If a rulepack limits Elite skills. then you might read that meta as meaning there will be fewer instances of Dodge and therefore don’t gain as much value from Tackle. However, someone else might look at that same rulepack, come to the same decision, and decide that this is a good tournament to take Zons, for exactly the same reason…. there’ll be less Tackle. Play and counter-play.
A robust pick might be to take Tackle anyway. Or it might be to roster Frenzy (poor man’s Tackle) as a compromise.
So, before adopting a counter-meta strat you need to answer:
How accurate do I think my read of the field is?
A strategy can become worse simply because too many people discover it.
Stabilising Metas and data mining
When more games played under the same ruleset, the meta first matures and then largely stabilises.
The Eurobowl and World cup rule sets are good exemplars here; they get used across multiple tournaments in the months leading up to the event. More and more games = more and more data points that are available to all. Usually available in Tableau from Mike Davies (Sann). Link
The extra data can allow, in certain contexts, a more accurate read of the meta. Similarly a very narrow, constrained, ruleset where some teams are perceived as non viable, can give you a more accurate read.*
*beware that the meta might stabilise differently locally
HE, DE, Necro, Orc, Human, Nurgle are popular
Stats above are from this page, which is great – Eurobowl Stats | BBTV
^^one of the best things about these sorts of data sets is that I can sit looking at them at work and people idly walking by think I’m analysing RNAsequencing data. Glorious.
Jimmy Fantastic’s Mega Meta Table
By a lucky coincidence Jimmy has done a deep dive into the current meta standings released at the same time I posted this. He’s generated the wonderful table above showing win rate for different races per game played. In his video he talks more about why the teams toward the top left are where they are. Thoroughly recommend watching on the youtubes – you can find it here:
Example – using EB rules meta data to make Exploitative picks
By popularity, High Elves are miles ahead of everyone else, then Dark Elf, Necro, Orc, Human. So, to identify an anti-meta pick we might ask “what’s good vs these 5 races” rather than simply “what’s good”.
Unsurprisingly the popular teams are also doing well, in particular HE, DE and Humans all have good win rates against most opponents. BUT how will you personally get on in a mirror vs a strong opponent while playing for the title?
To go anti-meta using the data, what we are looking for is teams that have high win rates (we don’t care about draws) vs ideally 3+ of these teams.
For example Elven Union have 50%+ win rate vs all 5 of the popular races (they actually have the highest win rate but are the 9th most popular race).
That’s partly why I took EU to Davo bowl a few weeks ago.
Top-level coach insight – Mike Davies (Sann0638)
If you don’t know Mike (or know his work) you probably haven’t been playing very long…. former NAF president, curator of all things stats, winner of lifetime contributor award, and a casual 867 NAF tournament games, 15 tournament wins and 40 organised tournaments.
I asked him to comment on this post before it went live and he made an important point:
“I wonder if there is something in there about teams having high general win rates but not tournament win rates? E.g. used to be Dwarves, maybe Orcs, now probably Nurgle, that go 4-2-0 fairly regularly but less often will go 5-1-0+ which is needed to win.”
This point is important and reflects the danger of interpreting stats out of context. Using the analogy from above, the win rate or non-loss rate is the peak in the outcome distribution, but often what we want to know is the width or the tail.
Improving your Play
Final section. One of the focuses that is worth having is understanding how to get better and move toward the elite.
Specialising your way to Victory
Experience has value even when the team choice is contextually sub-optimal
Operating a roster you thoroughly understand through hundreds of reps often yields better real-world results than switching to a theoretically “better” option you cannot navigate as confidently.
An average coach with 100 games may outperform their theoretical “better choice” simply because they understand the decision tree better.
So, an alternative or complementary strategy to boosting your tournament win likelihood is to become a race-expert.
In game theory this starts touching on bounded rationality. The theoretically best strategy isn’t always the practically best strategy for a particular player.
Game Theory Terminology Explained
Bounded rationality is the idea that people (or agents) make decisions using limited information, limited time, and limited cognitive capacity. Instead of perfectly optimising, they “satisfice”—they choose an option that is good enough given the constraints they face. For BB this is usually time and money, one or both likely limit your chance to go to a tournament every weekend. Or, possibly more accurately, the views of your spouse on how the time and money should be spent!
The good thing, of course, is with online play you can get plenty of reps in with your chosen race before ever putting them on the tabletop.
To win, you should play something you understand better than your opponents understand how to play against.
We mentioned Jairo above as one of the 5-tournament winners this year. All five were with Humans, but also 125 of his 175 tournament games are with Humans. Humans are an on-meta pick, but being piloted by a coach who is good with them also makes the most of their playstyle.
Not being a race expert is (likely) partly of why I went 4-0-2 at Davobowl a few weeks ago! Still plenty to learn.
Top Coach Level Insight – AndyDavo
Andy is known to everyone in the BB community. In addition to winning everything himself, he also coaches players to improve at their game. He gave me some valuable insight here into how those sessions go:
I have helped lots of people look at tournament lists, I often find myself getting people to do a mixture of two of the points you raise. We focus on looking for a handful of teams that are resilient and effective in most tournament rulesets – finding teams that are right for them. We then focus on getting some serious games in so that they are masters of a small selection of races. Next, we then focus in on specific events where those races are then going to be most effective. I often steer away from variance – while it’s likely right most of the time, I think it should be an option in my arsenal.
I’m thinking now it would have been quicker to have some Davo coaching than writing a 10-part game theory series! 😉
Winning tomorrow’s tournament or winning more tournaments in general?
I’ve touched on reflection and practice throughout this series and these concepts hold true for looking at our roster and how we performed with it.
I think we can break the outcomes into four categories, each of which tell you something:
Correct roster, not right for you. This leads to reflection questions about why it’s not right for you, and/or whether it is not right for you yet. Is it a skill issue or perhaps it doesn’t suit your intrinsic risk profile or preferred playstyle?
Sub optimal roster, but right for you. The reflection here would be why does this work, and whether this stays the best option for you as you get better as a player or more experienced with the team. Will playing a more robust roster help you to improve? Or, does this race/roster combo work for your personal risk profile.
Sub optimal roster AND not right for you. This might look the same as above but is a big difference. This is about recognising why you lost (or that you won just by luck).
Correct roster and right for you. Sweet spot, but again, needs good levels of self-understanding to correctly identify.
It is difficult to tell these apart… and partly because of outcome bias. So, let’s go there next.
The Danger of Outcome Bias
I’ll keep banging this drum; we shouldn’t judge decisions in a high variance environment based on the outcome.
One of the easiest mistakes to make is assuming successful outcomes validate decisions. A coach takes an unusual off-meta roster. They win a three-game event, or even just do better than anticipated. The roster is declared brilliant. But success alone does not prove correctness. The same strategy may have failed in ten parallel universes.
Whenever an off-meta strategy works, we should ask:
- Was the idea good?
- Were the matchups favourable?
- Did we get lucky?
- Would the strategy still look good after fifty games?
The answer may genuinely be “yes”. The important thing is asking the question. [Note that the same questions are valuable if it didn’t work too… but here you are likely to be reflecting anyway rather than basking in victory].
Why bring it up again? Well innovative thinking is valuable, but overconfidence is dangerous and the difficulty is telling the difference.
Humans are extremely good at building stories around successful outcomes. We are much less good at estimating how much evidence we actually possess.
A three-round event generates confidence very quickly. It generates knowledge much more slowly. Good metacognition* means constantly evaluating not only our conclusions but also our certainty in those conclusions.
*Metacognition is thinking about your thinking. It involves recognising not only what conclusions you have reached, but how you reached them, how confident you should be in them, and what assumptions or biases may be influencing them.
I mentioned Dunning-Kruger (DK) effect in one of my earlier posts as something to try to avoid. I think in a meta-game analysis setting, DK becomes particularly relevant.
DK is a cognitive bias in which people with limited knowledge or experience in a domain tend to overestimate their level of competence because they lack the expertise required to recognise their own mistake. The DK effect is often described as “knowing enough to feel confident, but not yet enough to recognise what you’re missing”
The important thing to note is that this is true even if one generally accurately knows their relative ability (“I’m not very experienced… but I know this works”). The problem is the lack of metacognitive ability to correctly identify where the bounds of the knowledge lies.
I encounter this frequently when teaching medical students. A student may leave my lecture feeling that they understand a topic completely (if they didn’t sleep through my droning on!). Months later, after usually (sadly) a failed exam, they realise how much complexity they missed by overestimating their understanding.
Blood Bowl presents a similar challenge. A coach may take an unusual roster build to a tournament and achieve a good result. The experience naturally encourages confidence in the strategy. However, the tournament result alone does not reveal whether the build was genuinely optimal, whether it correctly exploited a specific set of matchups or playstyle weaknesses, or whether variance played a larger role in outcome. Success answers the question “can this work?” It does not automatically answer the question “is this the best approach?”
The metacognitive challenge is resisting the temptation to treat a successful outcome as proof that further improvement is impossible.
Coaches who improve most rapidly are often those willing to separate evidence from certainty and remain open to the possibility that a strategy can be successful but can also still be refined
Steps to Improving
I also asked Scouseboy to comment – his full quote is below, but this part fits nicely in here:
My advice is simple: look at the results, the wins, the trophies, as a symptom. A symptom of applying the correct approach to the game, a symptom of placing yourself in the best mindset, with the best preparation, with the greatest chance for success. THAT’S how you measure success… by going in, eyes wide, bolstered by the knowledge that you’ve done all you can to bring those prizes home, but content in the ineffable truth that the pows and the skulls don’t actually matter at all.
The joy of the game is in the build-up, the anticipation. It sounds trite, but the optimising how you play is the real metric for success. Counting the trophies is just notching the bedpost.
Practice
Breaking the format of previous posts, we need to start with practice first.
You won’t get accurate, valuable outcomes by defining your objectives and hypotheses after the event… it needs to be done before* to be meaningful.
Whenever preparing for an event, explicitly define what form of success you will optimise for before selecting a roster.
Are you trying to optimise:
- Your chance of finishing first?
- Your average tournament placing?
- Not losing?
- NAF ranking points?
- Your own development as a coach?
- The opportunity to test a new idea?
- Win most TDs or most CAS or best Stunty or Best Painted.
- Ticking off a race and just doing the best you can?
Having a clearly defined objective makes it easier to evaluate decisions honestly afterwards and improves the decision making beforehand.
*I travel to tournaments regularly with guys from the Liverpool club, we always have this conversation pre-tournament on our journey across the country.
Then evaluate the meta against your adjective. Gather whatever data you have available. Decide what the meta-picks look like and assign a level of confidence to those picks. Then decide whether there is an anti-meta pick that might give you and edge and which you feel comfortable playing.
Reflection
During reflection ask:
- Was my read of the meta accurate?
- Did my roster choice match my objective?
- Did I correctly balance robustness vs volatility and exploitation?
- If I replayed the event twenty times, would I expect the same choice to perform well relative to my objectives?
Remember a strategy designed to win a three-round tournament may look poor when judged by average finishes, while a strategy designed for consistent rankings may not produce tournament victories despite being fundamentally sound. Re-aligning your expectations will increase your chance of going home happy!
Also, aim to consciously consider whether your conclusions are being driven by outcomes rather than evidence. If an unusual strategy succeeded, ask whether you would still believe in it if it had finished mid-table. Equally, if a strategy failed, ask whether the underlying reasoning remains sound. Value the feedback and insight of your opponents.
Top-Level Coaches Comments
BBJock – Team Scotland Captain
I have always been competitive when playing Blood Bowl and I do love the deep dives you go into with your analysis. However for me, your opening in this blog is the most important point. Having a good time, and ensuring your best efforts to ensure the coach opposite you has a good time means we’re all winners.
KFoged, is undoubtedly one of the best coaches in the world and I’ve played him three times across the board. I look forward to it because it will be an opportunity to test myself against the best, but what I really love from those games is Kare is the master of making sure you have a good time. I learn undoubtedly, but I laugh loads and enjoy every minute of that match.
So I’d encourage coaches to always be learning, and reading your blog will help, but really learning how to build a fun experience together will have coaches come back for more.
Scouseboy (Craig), winner of 3 tournaments so far this year (2 of which I was at), and perennially in and around the top tables.
I used to play Magic: The Gathering at the highest level. Multiple Pro Tours, Grand Prix Top 8, National Championship win, and so on. While some of the skillset transferred over to Blood Bowl, it’s the more esoteric stuff that sticks in the mind and resonates even now, when my weapon of choice is a dice and not a card.
One hoary adage that still rings true is this: twenty-five percent of the games you play, you’ll just win. You’ll roll sixes and pows, while your opponent rolls ones and skulls. On the other side, twenty-five percent of the games you play, you’ll just lose. It’ll be your turn to be screwed by the laws of random chance. After all, some days you’re the statue, and some days you’re the pigeon.
How you perform in the remaining 50% of the games you play? That’s what determines how good you really are.
On one level, that’s pretty straightforward. If you do well in the closer games, you’ll win more often, and the trophies will come. But on another level, it’s more nuanced. While the numbers are not scientific (although I’d argue they are close enough for poetry), the truth behind them is evident: some of the time, likely half of the time, you’ll be unhappy with the results. So basing your mood, or your success, or even your worth, on such a nebulous and ephemeral “prize” as winning is a sure fire slippery slope. If everyone in a fifty-coach tournament prized winning above everything, then there’d be forty-nine pissed-off people at the end of Round Three.
My advice is simple: look at the results, the wins, the trophies, as a symptom. A symptom of applying the correct approach to the game, a symptom of placing yourself in the best mindset, with the best preparation, with the greatest chance for success. THAT’S how you measure success… by going in, eyes wide, bolstered by the knowledge that you’ve done all you can to bring those prizes home, but content in the ineffable truth that the pows and the skulls don’t actually matter at all.
The joy of the game is in the build-up, the anticipation. It sounds trite, but the optimising how you play is the real metric for success. Counting the trophies is just notching the bedpost.
That’s why articles such as this, and Camelchops’ great series in full, are so useful, and so joyful. Because they give you a way to win the whole damn thing before you get to the venue.
And if all this fails? Just roll more pows. That should do the trick.
Previous Game Theory posts
Right, it’s been a journey, but that’s the end of where I wanted to go with the Game Theory Series.
I hope you’ve enjoyed it. It genuinely ballooned quite a bit compared with where I started and I think they improved along the way.
I might tidy up and condense everything sometime in the future. But for now, I’m gong to continue to try putting more of the concepts into practice.
Comments always welcomed.
Leave a Reply