I am back again with the next post in the Game Mechanics Series. This post took a while to write because, unfortunately, I had to manually collect all of this data as I couldn't find anywhere this information is captured. The topic of today's post is to examining how the rating system interplays with your day-to-day playing experience. If you missed the last post, I'll be linking that down below.
Post 1: Unpacking the Splinterlands Rating System
Rating Point Inflation and How it Affects You
As we described in the last post, one of the primary core mechanics of the rating system in Splinterlands is the use of inflation mechanics. In a pure ELO based system, it is nearly a zero sum game. The only way to introduce new points into a closed system like a pure ELO is for a new player to start (points are removed from the system when a player retires). However, as discussed in Splinterlands, there is an intentional inflationary aspect to the game as they have a tiered reward structure. Without this inflation / reset system (season), a player who gains tons of points can simply never play and keep reaping the rewards should no one be able to catch him (other games use deflationary mechanics like rating decay, which is not used in Splinterlands). The inflation rate is designed such that it attempts to sort players with their peers, and provide enough difference such that stronger players won't continuously bully weaker players (better gameplay experience)
Inflation is primarily injected two ways:
1. Stronger players beating weaker players (>200 point difference in rating)
| Delta | Expected | Actual | Inflation |
|---|---|---|---|
| 200 | 10 | 10 | 0 |
| 240 | 8 | 10 | 2 |
| 300 | 5 | 10 | 5 |
| 360 | 2 | 10 | 8 |
| 399 | 0 | 10 | 10 |
[Source: See previous post here]
When stronger players beat players they are significantly higher rated than, the winning player gains more points than mathematically they should, generating point inflation.
2. New players adding points to the pool
[Source: peakmonsters.com/dashboard]
Over the last month, around 5500 new players are starting each day. In Novice any battler with zero points will grant points to a winner, even though they lose none (as they have none to lose). Thus, points are generated and added to the ecosystem each time a zero point person battles. It's a bit hard to quantify how many points this adds per season, but with that many new players each day, the effect is not negligible.
How Inflation Manifests Each Season
I tracked the rating of each tier from bronze to champion every day of the season for a few seasons. If you look in the top left corner of the battle screen, you will see a time remaining in the season. Graphically, this data is shown as a function of the time remaining versus the rating of different players on the leaderboard at a fixed slot (I arbitrarily chose to track the 1st, 25th, 50th, and 100th place for each league). Zero days remaining data is based on the end of season result (shown in the last season leaderboard). Without further ado, here is what I found:
Bronze:
Silver:
Gold:
Diamond:
Champion:
As you can see, all of these graphs look very similar. Each day the rating goes up in a very linear fashion (R2 of this data was generally around 0.96 or better). What you may also notice is that the slope of each position is similar, but the higher ranked players generally gain a bit of separation over time. Numerically, these positions tended to gain around the following number of points per day:
Players who are ranked at the top of the league tended to outpace lower positions in the league, with champion being the exception. Champion is a bit of an outlier as you literally have no one above you to farm the rating points from. In order to keep climbing, a win rate of ~80%+ is needed (roughly you risk losing 36 points versus a max 10 point gain, so you need to be about 3.6:1 favored to break even).
But what does it mean?
Indeed, what benefit can you gain from this information? Remember, you get reset each season based upon your rating last season. The reset values are as follows:
| Final League Finish | Next Season Reset Rating |
|---|---|
| Bronze III | 0 |
| Bronze II | 100 |
| Bronze I | 300 |
| Silver III | 500 |
| Silver II | 700 |
| Silver I | 900 |
| Gold III | 1200 |
| Gold II | 1400 |
| Gold I | 1600 |
| Diamond III | 1900 |
| Diamond II | 2200 |
| Diamond I | 2500 |
| Champion III | 2800 |
| Champion II | 3100 |
| Champion I | 3400 |
1. How many loot chests you can expect to earn in a season (i.e. budgeting)
By extrapolation, you can roughly figure out your "true" position by what rank/rating you ended when the season closes. Warning: Maths ahead For example, I generally finish around 4000 rating each season (I have ended here +/- 100 points for the last 5 seasons). I will get reset to 2800 each season. The season is roughly 15 days long, meaning I will get 15 days worth of quest chests assuming I complete the quest daily.
| Tier when Claiming Reward | Daily Loot Chests Reward |
|---|---|
| Bronze III | 1 |
| Bronze II | 1 |
| Bronze I | 1 |
| Silver III | 2 |
| Silver II | 3 |
| Silver I | 4 |
| Gold III | 5 |
| Gold II | 7 |
| Gold I | 8 |
| Diamond III | 10 |
| Diamond II | 12 |
| Diamond I | 14 |
| Champion III | 16 |
| Champion II | 18 |
| Champion I | 20 |
Looking at the charts, a diamond/champ III player like myself will typically finish with a rating of about 4000. Comparing that against the chart, it expects a bottom leaderboard person to average gaining 85 pts/ day. Starting at 2800, over 15 days, that would estimate me at ~4075 pts. Not too far off. While this is not exactly linear (forcing quests and win rate variance play a big role), I can roughly use this info to gauge where I will be on any given day of the season.
Using the chart above, I estimate I gain 80 pts/day (rounding down a bit since I generally don't end up on the leaderboard, just slightly lower), I will spend:
3-4 days in diamond 3, 3-4 days in diamond 2,3-4 days in diamond 1, and then the remainder of the season in champion 3. Thus, I can estimate my total daily loot chest to be:
4 days x 10 = 40 chests
3 days x 12 = 36 chests
3 days x 14 = 42 chests
5 days x 16 = 80 chests
Total: 158 chests
Combined with the season end bonus (80 chests) I can roughly gauge how much I can expect to earn in a given season from rewards. Splintercards has a wonderful tool for estimating the EV of loot chest. This info can be useful, especially for people who are renters.
The tool roughly estimates that each chest is worth around 49 DEC at the time of this writing. With the total of 238 chests I expect to open, that would give earnings roughly equal to 11,662 DEC / season.
If I plan to rent, this number should be plugged in as part of budget setting for rental cost/day. Assuming you want to at least break even, then this figure can be used to say that I should be spending less 777 DEC /day. This neglects your DEC earned along the way from wins (there are ways of estimating that too, but that's a whole different post), so all of that would be considered net gain to be saved to actually purchase cards for future use. This can be adjusted for any league/CP that fits your situation.
2. Maximizing gain potential if you are renter
So you want to climb fast while minimizing expenses? One of the ways to do that is to actually skip playing days. Essentially, your true skill level stays roughly the same (if you are new/get new cards/etc. skill level might change), but your rating will reset. Points funnel upwards from the lower ranks to the upper ranks, so by not playing you are going to be routinely matched with players who are likely lower skilled/carded than you. The more days you wait, the more mis-ranked you will be and the more likely you are to rank up while taking advantage of win streaks. This will allow for easier quest completions, higher win rates, and better return on your ECR. Higher DEC pools are available for higher rated battles, so getting to those higher ratings is very lucrative assuming you can obtain the necessary collection power.
There are many strategies for this, but rental costs tend to climb very slightly over the season, then peak the last 24-36 hours. Obviously there are many goals in this game, but one potential strategy is to get to champion III to obtain champion points. Champion points allow you earn air drop points (SPS) passively. If this is the primary (only) goal, then what you can do it basically play only 1 day in the season. If you wait around ~9-11 days into the season, climbing from 2800 to 3700 should be entirely possible with a high win rate while only needing to use 1 day's worth of rentals (this is basically assuming you know/have the skill to be low end top 100 with appropriate cards). Then basically use up the rest of your budget on the last day to lock in collection power.
3. Keeping Higher Win Rates
This is a corollary to the point above, but remember you only win DEC, when you win. The first few days of the season are often blood baths, as the stratification between the higher rankers and lower rankers is low. As they higher rankers keep winning, they move up and you don't face them any longer (or much rarer at least). By waiting a few days into the season, you can ensure you fight against people more your level (or lower), which can increase your win rate, and thus your DEC earn rate. Win streaks also provide bonuses (up to +50% DEC rates) when you keep winning, so there are definitely incentives to taking advantage of better matchups when you can. There is a fine line between ECR/win rate/base DEC earning which I will examine in a future post.
Wrapping up
I began this research mostly out of curiosity for myself, as I try to maximize my own play. One of the goals I wanted to investigate was to track the consistency of the ratings throughout the season and how the interplay between playing for leaderboard (diamond) was versus collecting extra chests and airdrop points from being in Champion. Another aspect was whether I should personally rent for better cards during the season (ideally a better win rate), and whether the extra earnings from a higher win rate would be worth it. This post is really just an introductory teaser to that discussion. Look forward to the future posts where I delve deeper into the world of min/maxing win rate vs. earnings!
Congratulations @rooo! You have completed the following achievement on the Hive blockchain and have been rewarded with new badge(s):
Your next target is to reach 3000 upvotes.
You can view your badges on your board and compare yourself to others in the Ranking
If you no longer want to receive notifications, reply to this comment with the word
STOPTo support your work, I also upvoted your post!
Check out the last post from @hivebuzz:
Thanks for sharing! - @alokkumar121
Congratulations @rooo! You received a personal badge!
You can view your badges on your board and compare yourself to others in the Ranking
Support the HiveBuzz project. Vote for our proposal!