Draft, from my perspective

I’m really happy with my team. I had a couple possible “paths” I wanted to take with the draft, and this was one of them. I basically got exactly who I wanted. (The other path was getting polo.)

I knew that being too rigid in this auction format could backfire, so I was prepared to be a little flexible. Observers did see me (up)bidding on a variety of players, but also saw me waiting for my moment.

I’m going to touch on numerous points, and I think the best way to organize it is chronologically. (warning: long and rambling)

Here’s how I approached the draft.

I fully intend to be done with dota after this rd2l season. This emboldens me to share my thought process, whereas if I were to play future seasons, I might be reticent. It also informed my draft approach. I wanted to get people I wanted to play with. Drafting “correctly” for competitiveness, just isn’t “me” anymore. Ambition and strife seem so unnecessary, these days.

So if I wanted to play with polo, it was either now or never. I’ve been an admirer for many years. There are two “polos” in rd2l. The other one is one of the top bulldogs of the melvs. That group always liked to find conflicts with me. There are types who cannot understand, and mock what they cannot understand.

And then this polo is so gentle and radiant and beautiful. Like golden hour in the quiet mountain-side.

But still, there must be a limit to how much I can bid. Clearly, the right number was somewhere less than all of my money.

~~ Draft Prep ~~

How to decide on the bidding limit for a player?

I decided to use something from game theory: The principle of indifference. The point where I am indifferent between my getting a player, and my opponent getting him, is the correct bidding limit.

How do you find the “intrinsic value” of a player?

Understand this deeply: Once captains are assigned money, all we care about is getting the most value out of that money. It does not matter who the captain is, and what formula is used to distribute the money.

Once the money is in hand, all that matters is what value (mmr) the money can get. It is BLIND to whoever the captain is, and howsoever the money was given.

So the total money available (all captain $ summed), corresponds to the total mmr (value) available to capture. The lowest mmr player is worth nothing, because he will be free.

With this, you can calculate the “intrinsic monetary value” of the players. This is the “greg” model used in the early auction seasons. It looks something like this:

It is this value that I’ve referred to in previous blogs. Each dollar $ corresponds to about 30 mmr.

So if you buy players below their intrinsic value, you’re winning. If you pay more, you’re losing. Simple?

It’s complicated by many factors:

– You must draft exactly 4 players. This is consequentially the same as saying “leftover money isn’t worth anything,” but people who don’t understand only know how to mock.
– Role distribution, as my last blog talked about
– Higher mmrs may be worth more, cores may be worth more, etc.

___

So with $550 to spend, as you can see, if I draft Timber Hearth (adjusted down to 4500 on the draft sheet), 1 or netflix second monitor, stonks, and 1 more, I’m getting close to around my rightful value. If everyone overpays for all the higher mmrs, and then I scoop up 4 players around that level, I can’t really be too unhappy.

One reason I blogged so much, and vibe checked people on discord, is to find people who might be compatible. The players I got were literally the ones I was targeting.

My other plan was if I got polo. Then I would’ve tried to get a value carry like george, fallen, or fulcrum, and played offlane myself. Or I would’ve tried to still get Timber Hearth or scrub, and fill with lower mmr supports like Buck.

___

According to the intrinsic value model, how much mmr could each captain expect to capture?

• Buck’s Scaled is the amount of starting money captains would have had, going by intrinsic mmr/greg’s model.
• Crub Cents is the actual starting amounts this season
• Implied Team Avg is the team average mmr that each captain could expect, if the auction runs according to intrinsic values.

Why does the auction deviate from the simple mmr model?
1. Higher mmrs rightfully go for more than the mmr model implies. This effect is, in my estimation, over-corrected for by captains. This creates “mistakes”, or “inefficiencies”
2. It is harder for low mmr captains to benefit from these drafting “mistakes,” because they have too much money, and are thus forced to also make the drafting mistakes, in order to spend out.
3. Higher mmr captains are in a position to capture more than their rightful value, and low mmr captains cannot capture their full value.

___

Opinions

1. I think it’s correct to adjust the money, giving more to lower mmrs. I think it’s inequitably over-tuned.
2. Using previous season bidding, which you have no reason to believe is correct, is odd. If you inject more money into the system, why wouldn’t you just expect it to inflate the incorrect-ness?
   And then you want to, by fiat, set the order of the nominations, and display the “correct” values to draft players at, for which there is no solid game-theoretic reason to believe are correct.
   (Are you following? You derived the values from previous season behavior, and then you want to enforce that behavior on people in future seasons)

~~~ Draft Day ~~~

This is really long and rambling. If there are specific questions you’d like to address, please ask and I can answer in a follow-up.

So, I was actually planning to nominate stonks first, planning to take him. I thought that captains had a tendency to just overvalue high mmr players, and not necessarily all early players. I thought that captains would be reluctant to purchase a high mmr support, so early, because it would cripple their further purchasing ability. (Which I think is correct, by the way.) I might get stonks at a bargain.

But the draft, surprisingly, was jumping all over the place, and not starting at high mmrs. I saw that people were just overvaluing people across the board. So I didn’t nominate anyone I actually wanted, and just waited to scoop them later.

I was playing with fire. There were several possible problems.

1. All the high mmrs would be gone
2. I might not actually be able to get who I wanted, because two other low mmr captains, Nifuji and Mantis, and Waifu and Gronaldo, and Vuvu, and others, still had quite a lot of money.
3. Then, I would be stuck overpaying for people anyway, and having no one good left to spend money on.

If someone wanted to screw me (or was just dumb), they could outbid me using their money. For example, let’s say vuvu is hell-bent on taking a player from me, using all his money. He will be screwed, because he has no money left, but I will be screwed, because I needed that player. He screws us both.

It was a real possibility, that didn’t happen. Partially, I got lucky. It’s still safer to overpay earlier, and get rid of some money, when you are a low mmr captain.

___

I did have scrub’s valuations in my draft sheet, as a reference. I knew that high mmr players should go for more than mmr-implied, and scrub values seemed like a reasonable sanity check. (Except it gets unreasonable at the top. 381 for hobo is mathematically ridiculous. Keep in mind the scrub model also values the lowest mmr players at negative money. It’s just a model.)

I did not “bid up” anyone just for the sake of gouging people. I bid up to the point where I was indifferent. If my opponent stopped bidding, I was fine with it.

I was fully intending to overpay for polo, justifying it with the Crub welfare money I got. But toth follows his heart more than I do, wcid.

By the time it got to Eli, I was really starting to feel like I had to spend money on anyone available. I probably could’ve gone higher, and zombies probably would’ve too. But I wasn’t bidding just to bid anyone up… I didn’t really want Eli for compatibility reasons.

If you think 315 for hobo was cheap… then why was I the only one bidding on him with redground? I also probably could’ve gone higher here, and redground probably would’ve too, but it was still relatively early and I thought I’d have other opportunities.

I found it useful to think about the possible money-player distributions. Like 350 on polo, 100 on stonks, leaving me with 100 for a value carry + low mmr support. It was kind of a sanity check on what I could possibly get.

Practically, if I did get polo, it would’ve been better for me to spend my remaining money on a 3rd core instead of putting it into supports.. but I really wanted stonks 🙂

For offlane, I was pretty set on either Timber Hearth or scrub. If one of the higher mmr players fell into my lap, it would be fine, but I wasn’t expecting it. I got the feeling scrub was kind of either/or on playing with me. He seems to be kind of an uncommittable person in general. So I didn’t really plan on bidding very much for him. Luckily, he was nominated before Timber Hearth.

On treebeard’s draft stream, scrub was saying how he “wouldn’t let [my] team win games” if he were on my team. /facepalm

If you would’ve loved me, I was prepared to love you!

GL HF with midorchen and smack.

___

IDK there’s lots of other small details, but it seems disorganized and rambling. Ask me if you have specific questions.