StarCraft: Remastered Developer Update 3 - Page 4

On December 24 2017 02:31 hyfrehyfre wrote:



Go check maps like "EDDY" they have amazing elongated ramps... i think they are going for something similar. I Hardly understand what they mean with "ramps" to be honest.

On December 24 2017 05:49 TT1 wrote:


You can have the 1st rank up be like 100 points, past the 1st rank move it up by increments of 300 points. That's how Fish did it. It took 4 wins to rank up from F to E (first rank up), everyone started at 1000 pts and they'd get 25 pts for an evenly ranked win (for example F vs F).

Assuming our base rating starts at 1500 like it does now, it would be: 1500 > 1599 (F), 1600 > 1899 (E), 1900 > 2199 (D), 2200 > 2499 (C), 2500 > 2799 (B), 2800 > 3099 (A), 3100+ (S). We could extend A rank from like 2800 to 3299 (instead of 2800 to 3099) to make S rank feel more prestigious, S rank would then start at 3300.

source: http://wiki.teamliquid.net/starcraft/Fish_Server

Current system (effective April 8, 2017)

Rank Points
S >2600
A 2300–2599
B 2000–2299
C 1700–1999
D 1400–1699
E 1100–1399
F 801–1099 (everyone started at 1000, took 4 wins to reach 1100/E rank)
F- 1–800

So for the current ladder system the equivalent would be to start at 1500 (F) and rank up at 1600 (E), aka 4 wins. F rank (base rank) would have to be around 1300 - 1599, F- would be 1000 to 1299. 0 to 999 would be the beginner rank, you could split it up into 2 ranks if you wanted to, say 0 to 499 and 500 to 999. Call it w/e rank you want.

On December 24 2017 08:47 Excalibur_Z wrote:


I don't entirely agree. That's very close to the "traditional" concept of a ladder. Super grindy, everyone condensed tightly into one region until chunks of players start branching out. Eventually subchunks break out, then subchunks of subchunks, and eventually you have crazy population numbers where each rank has like 5% of the population of the rank below it. It ends up looking like a pyramid.

I'm of two minds on this. When I played SF4 and Soulcalibur 5, they had grindy ranked play like this (Japanese game developers love the hyper-grindy rank climb). I hated it. Sure it was prestigious to see (or be) an A-ranked player, but it took literally thousands of games to get there. Elo moves about as slowly. You never want to have to say "yeah Flash is B, and there are only 5 B-ranks out there right now, but eventually he'll reach A or S", it should be evident almost immediately who is the best player in the world. Setting up "aspirational" ranks only artificially delays what the disitribution should look like.

One thing that's really good about the SC2 system is that you have equal distance to rise or fall from the median starting point. It also moves a lot more quickly than Elo. However, one of the main reasons why the SC2 system works is because you're only allowed a single account per region. In SCR you can create unlimited accounts, which can lead to inflation because if you go from 1000 to 900, you can always just make a new account at 1000 and climb rather than use your first account. It would probably still be better than the pyramid-shaped supergrind though.

On December 25 2017 03:43 Leeoku wrote:
Changing the entire BW history, wider ramps

On December 24 2017 13:32 Excalibur_Z wrote:
I'm a little biased obviously but I think the ShieldBattery ladder approach is best. That had I believe 11 buckets and you started in the 7th highest one, so almost equal room to grow in either direction (I was taking into account people who would just quit or make new profiles once they got too far below the starting value). Each bucket was equal in rating size, but were intended to target certain percentiles, similar to the Heart of the Swarm ladder distribution. Every so often we would round the values off and recapture those percentiles if they drifted from the intended targets.

My SB design was still using Elo but with some modern adaptations:
- larger K-factor if you have less than a certain number of games played
- after you cross that number of games played, your K-factor would decrease if the outcome of the match was expected (your predicted win chance by rating difference was less than 50% and you lost, or more than 50% and you won)
- your K-factor would gradually increase with a streak of unexpected outcomes

It was kind of an attempt at approximating the TrueSkillesque SC2 model, but Blizzard can do that much better than I can. But the basic gist is:
1. Rating ranges that are fixed
2. Equal(ish) opportunity to rise and fall from the origin
3. Rapid and aggressive initial placement

The season roll thing wouldn't have to reset MMR. In fact, there's no reason to ever reset MMR because it means you're willfully discarding millions of data points that have led to player skill definitions. I wouldn't expect the season roll to change that at all. The only thing the season roll will do is change the map pool and adjust the league boundaries if needed (that's the plan with SB also).

On December 25 2017 06:20 castleeMg wrote:
having alot of problems with matchmaking, getting matched then reset to the "searching 1v1" lobby screen, happens like 10 times b4 i get a game, not even exaggerating. is anyone else experiencing this problem? ive seen a post on the starcraft blizzard forums about it but no response from blizz yet.. im wondering if its a port issue but its geting super annoying

On December 25 2017 08:54 Alejandrisha wrote:
i'm thinking 2v2 ranked is never going to happen without ~5 minute queue times. they waited too long

On December 25 2017 08:47 HaN- wrote:
I'm curious how the team match making will handle the inevitable small pool of players.
Blizzard won't be able to make the queue times short and guarantee players of similar levels at the same time.

This is even more an issue because 2v2 players from my experience in public games are either very bad (think D-) or very good (think A+). You'll have a mix of players who have been mastering 2v2 for 10+ years and new comers from the 1v1s scene who will need to learn 2v2 in and out from the begining.

If good and bad players run into each other too often, it won't be fun for anyone.

This was the experience on iccup in the last years..
Maybe that now with the global ladder, having koreans can fix these problems but then what about the latency.

On December 25 2017 08:54 Alejandrisha wrote:
i'm thinking 2v2 ranked is never going to happen without ~5 minute queue times. they waited too long