Sierkovitz

7 Feb 2025

Since I got a load of new followers who, I suspect, have no idea who I am, a short introductory thread. Starting with the basics: I look at data in Magic and try to convert numbers to actionable advice. You might have heard me on Limited Resources, but I also have my own podcast, Magic Numbers. 1/8

Evolutionary biology has a concept that diversity begets diversity. Existing diversity creates niches, novel conditions and this is driving future diversity. This thread is about how early data impacts the creation of future data by changing player behaviour. 1/12

Surprising power level of Gollum boosted the WRs of early data consumers and top players. It let them take their easy first day wins, but late adopters reap smaller rewards. They miss out on the bonanza of first days. And the later you realised it, the smaller the reward. 11/

linktr.eesierkovitz Official: Twitch, Twitter | LinktreeM:tG data analysis. Magic Numbers Podcast, collaborator LR, mtgazone.com

Hopefully you liked this graph-based storytelling. If you did - why not support my merry ramblings. Buy me a coffee. Sub me on YT, or sub my podcast. Or hey - wild idea - support my content? You can find all you need in my linktree: linktr.ee/sierkovitz 12/12

There are other consequences of that. A strong card you can safely wheel is a superpower for its color pairs. In the early days, you can prioritise harder to get cards, and count on getting this key common late. This is what top players pray upon in 1st days of the format. 9/

But as format evolves, cards like Gollum compete with harder to get cards. If you pass it, it's unlikely to come back. And the WR suffers. From a card 4%p above average, Gollum turned to a 2%p above average - all driven by pick order changes of it and other B cards. 10/

Even at the late stages of the format, the lower WR players don't pick Gollum as often ad top players - partially because of under-valuing it, partially because they are more frequently pushed off top color pairs. Information gap is one of the reasons for WR disparity. 7/

And in part picking Gollum earlier is forced by the environment. As knowledge spreads, even (stressing it) even 17Lands non-users get the message. Here you see how the card goes from wheeling 40% of the time to 12%. This requires all Arena users to start picking it early. 8/

First - 17L users. When we see a card that overperforms expectations, they start noticing it. In case of Gollum, it was picked at 7.2th pick, and slowly got taken earlier and earlier. Only 4-7 days after format dropped it was already 5.9, and stabilised around 5. 5/

But not every 17L user is the same. High win rate players use 17L not only to log their games, they learn from it. They adjusted almost instantly. Lower win rate players are slower in their uptake. It can be they are less invested, don't consume content etc. 6/

Why Gollum? It was a card that we missed in early evaluation of the format. No set reviews thought much of it, yet it became one of the top commons and held that rank throughout its run on Arena. Misses happen, but it's the reaction to those misses that matters. 3/

And reactions will be different for different player groups. 17Lands users, top players, low win rate players, even - I know - they exist, I'm surprised myself - 17Lands non-users, all those groups will react slightly differently. And it's worth knowing how and why. 4/

I will be using Gollum, Silent Slinker to depict this. We are having a story in 4 graphs that show us how early data can impact future developments in the format - perfect while we are waiting for Start Trek to come out and we can experience this process 1st hand. 2/

Reminder: win rates on 17Lands are averages. So if Azorius has highest win rate, it means average WU deck is better than average other. Shortcutting it to WU > rest is a dangerous trend. It's easy to do so, but it's also incorrect. You play a deck, not an average of decks. 1/8

But the individual variance between WU decks' win rate is larger than the variance between color pair win rates. If you remember that and navigate your draft looking for most power, you will start winning more. Range of what you draft is a sign of a better limited player. 7/

To get that range you need to experiment and not be afraid of stepping out of your comfort zone. And also notice when you are shoehorned into a weak version of a good color pair. This is what you should be using your data for, not just checking the rankings of color pairs. 8/8

Top players win for many reasons. Good game play is one of them. But another is knowing when it is correct to draft a "weaker" color pair because it is open. And when it is open, you are going to get a version of it that can be better than the best average color pair. 5/

Conversely, when you try forcing a good color pair while it is not open, you end up with Cryotheory Adepts deck and the color pair advantage is gone. Knowing color pair power is important in drafts where nothing is particularly open, then color pair power can be a tiebreaker. 6/

But what if our Boros deck has Ajani? Well, WR decks with Ajani have an average win rate of 61% - head and shoulders over WU average. WR decks with Kiora are close to WU decks (still below though). Context matters. Decks are individuals not populations - treat them as such. 3/

Same goes for sub-par WU decks. If your Azorius deck has Cryotheory Adept in it, on average its will bo only marginally better than a WR deck. If you put blue never-skip-a-leg-day Yargle - your win rate plummets to 51.4%, worse than any color pair average. 4/

To visualise it, here are the color pairs in FRA listed by win rate. You see - Azorius (Fatehold?) is doing well. 56.6% win rate is above the rest. Not by much, but significantly. Importantly, it's 3%p above Boros, considered to be one of the worse decks. 2/

Morning train commute of my daughter, we are waiting on the platform. Announcer: "Train is cancelled due to leaves on the line". Passengers: "Rabble rabble unacceptable" Me and my daughter looking at each other: "Poor Uncle Iroh"

My idea for Magic promos: Spotlight: Fetch lands for the foreseeable future. 4 Spotlights reward one type of fetch, 16 Spotlights per year, so 4 fetches per year - we'd get 2.5 years of fetch promos. Normal for participation, foil SERIALISED for top 100 with spotlight name. 1/5

RCs: Each RC gets a set of basic lands typical for the region where RC takes place. Top 100 gets additional foil, serialised set with RC location and date. Lands are changed every year but for a year each RC in the region gets the same promos. Congrats Australia ;) 5/

Is it a perfect proposal? Probably not. All I want is to start to the discussion. Important aspect: winner and T8 promos are worth something and sought after. Spotlight and RC foil promos are serialised. A memento of good performance or a selling point. What do you think? 6/6

* I know UBs don't get SC. They should. I didn't put Start Trek as it would have to be cards from the set. But Marvel has enough of back catalog or can adapt Magic cards with franchises theme, which I chose to highlight with Bolt Bend. Also note - SC promos are color themed. 3/

Kamigawa: Titanbreach: 1st Ozolith, T8 Kodama of the West Tree, all: Tamiyo's Safekeeping Zhalfir: 1st: Solitude, T8: Grand Abolisher, all: Teferi, Time Raveler 4/

SC promos. Thematically linked. 1st - value commander textless, T8: Themed Value card, all: commander staple Nauctis: 1st: Force of Will, t8: Lord of Atlantis, all: Offer u can't refuse Darkhold Destiny*: 1st The Soul Stone, t8: Mjolnir, all: themed art Bolt Bend 2/

My latest Magic Numbers on early FRA data is out. This set is dope and dope sets make data tougher to interpret. No more straightforward looking up the win rate. But some interesting things pop out. Like G is a better support color you can splash in than base for 5C. 1/2

youtu.beMagic Numbers #195: Early 17Lands.com data for Reality FractureYouTube video by Sierkovitz

You can see all of it here or get it from your podcast feed. Which I think is very much worth it if you want to improve in limited. But I would say that, wouldn't I? youtu.be/W32TboJw7dY 2/2

In case you are wondering how busted is the R Ajani in FRA draft. I started measuring cross-format win rates in Midnight Hunt, by normalising win rates across sets. Since then only 34 cards of 7871 analysed got >60% WR. Zero cards got >63%, but Ajani is 65%. Outlier of outliers.

Today: evolutionary biology, ecology and draft metagames. A new format is upon us, so some thoughts on how to read an emerging limited metagame, and maybe metagames in general. Magic will appear, but first some biology, you can survive it - you are fit. 🧵 1/14

Biologists call this negative frequency-dependent selection: a variant's fitness drops as it becomes more common. This keeps rare traits exactly that - rare. Becoming popular is what breaks it, so naturally they revert back to being rare after successful period. 13/14

So next time you read "nobody's playing it and it's busted, how come?", in limited or in constructed: it's probably busted because nobody's playing it. The moment people start, that stops being true. 14/14

Biology knows this strategy. Some traits succeed only while rare. Example: New Zealand mud snails. Parasites evolve to infect whichever snail clone is most common, so common clones crash and rare ones rise, with a lag. Being common is what makes you a target. 11/14

Niche limited decks work the same way. Dovin's Acuity decks in RNA were great while unknown and close to undraftable once the secret was out: key cards stopped wheeling, cards that used to come for free got contested, and sure-thing drafts turned into trainwrecks. 12/14

This happens often but rarely gets communicated: the early data volume drowns the signal, and the narrative has already formed. Shrewd players know this and learned how to exploit this persistent gap format after format. 9/14

Our example BR is the best deck in a vacuum. But they understand, we're not in a vacuum. Their drafting becomes almost constructed-like: prepare a detailed plan and a skeleton for an underdrafted strategy, then draft it through the holes in the metagame, unopposed. 10/14

The reality is more complex. Around week 2 the metagame stabilises, and only then can you game it. But by then the narrative exists, and the top pairs from weeks 1-2 dominate the narrative. They are also drafted far more than on day 1. 7/14

That opens space for other strategies, not necessarily the ones with the most raw power: - we've had time to find the right cards for them - we know they're open - we know the weak spots of the power decks - the power decks now face competition in the draft 8/14

Early on we're all in the dark. We don't know how to build decks or what the terms of engagement are. This stage runs mostly on raw card power, general skill, and how well players carry lessons from previous formats into the new one. 5/14

The first week produces the most data: most games are played and logged on 17Lands then. It's a useful dataset, but I think it carries too much weight in the community, given its quality. If week 1 says BR is best, the post-mortem will likely read: "It was a BR format." 6/14

Our genomes record this. Immune gene variants (e.g. ERAP2) were favoured during the Black Death: if your village was dying, carrying one was a large advantage. Today some are linked to autoimmune disease, but the costs are small enough for them to prevail in populations. 3/14

Three ideas of evolutionary biology carry over to draft: selection, timing, opportunity. A draft format works like an ecosystem: the set defines what's possible, but so do pick orders and players' understanding of how to build decks. That's why formats evolve over time. 4/14

Evolution runs on diversity. Life keeps generating variants, and natural selection favours the ones that fit current conditions. Those spread. The rest don't vanish: they persist at low frequency at the margins, and when conditions change, their time can come. 2/14

New Magic Numbers are up on YT. Our eccentric, deck-centric set review looks at how my esteemed guest (see below for how to find them) would build decks for all 10 supported archetypes. Through that we learn about the set, synergies, removal data, etc. youtu.be/N99E4pfp0_w 1/4

discord.ggJoin the XMage Draft Historical Society Discord Server!A welcoming community for Limited Magic players of all skill levels to experience each set released throughout the history of the game! | 2099 members

Springbok is helping a community that organises old drafts online. XMage Draft Historical Society even has software to help playing the games. Miss drafting Tempest? Never tried Conflux? Check them out here: discord.gg/tpdSfgdAMc 3/4

metafy.ggReality Fracture: How to read early 17Lands.com data?About the Event Finding your way in the early format is key but also laden with traps. You can gain a short term advantage from abusing open god archetype, but often it leads to lower win rate in late...

Also, don't forget group coaching about analysing and interpreting early data I will be running on Thursday. metafy.gg/events/reali... 4/4

twitch.tvtherosemarytea - TwitchThe Patron Saint of MTG Mediocrity. She/her. Buy my spooky beats here https://sweetpermadeath.bandcamp.com/ — therosemarytea streams live on Twitch! Check out their videos, sign up to chat, and join t...

The Rosemary Tea is one of the most interesting drafters on the market, with great gameplay and incredible intellectual capacity to make any deck durdly. Did I mention a great eclectic taste in music? www.twitch.tv/therosemarytea 2/4