![Designer Diary: Only a Few Cards, or The Making of an Automa for Patchwork](https://cf.geekdo-images.com/d9efla2ZzBp9cf3O1RjEXA__medium/img/MX5OkSrCD6tqu8vPAf9oFN3aD1o=/fit-in/500x500/filters:no_upscale():strip_icc()/pic4388154.jpg)
The way those resources are interwoven is new, fresh, exciting, and brilliantly streamlined. In addition, you have to look for the right shapes and which ones your opponent might be going for. Furthermore, with careful planning you can manage to get two or three turns in a row. This is one of my most underestimated games.
After playing the app, I immediately bought a physical copy — but unfortunately I didn't find the right people to play the game with as often as I wanted. I wished someone would create a solo variant for it.
Fast forward to SPIEL '16: Cottage Garden was released, with this game having some similar concepts to Patchwork. Morten Monrad Pedersen (Automa Factory) bought a copy, and so did I. Some weeks later, Morten presented an Automa for it. I tried it, and it worked fine but didn't grab me. That was not Morten's fault. I just found Cottage Garden itself pretty unexciting. It was missing most of what I loved about Patchwork and quickly gathered dust on my shelf, so I wasn't too much interested in Morten's Automa version either.
Instead I (and our third Automa Factory member [David Studley]) tried to push Morten into adapting his Automa to Patchwork. Unfortunately, Morten neither had the game nor the time.
Cottage Garden to Patchwork
It looked like I had to do it myself. My first version adapted some of Morten's Cottage Garden concepts. Instead of placing Automa's patches on her quilt board, they were placed beside it. A marker on the quilt board counted the squares the patch would have covered. It was moved in reading order along the empty board and when it reached the end, Automa was considered to have fully filled her board.
Other than that, my first Automa behaved much like a human player: She gained and spent buttons, she passed when she couldn't afford a tile, and when she could afford more than one of the available tiles she made her choice using a fixed tiebreaker list. The special tile you get for completing a 7x7 area was placed on a specific space on Automa's quilt board, and when her marker reached it, Automa was considered to have created a full 7x7 area and claimed the tile for herself.
I played this a couple of times and sent my draft to Morten and David. They both showed interest, but were way too busy to invest more time in it. That said, Morten mentioned he might be interested in developing this and offering it to Lookout Games, the German publisher of Patchwork. He asked whether I'd be willing to wait until he had the time to join in. Oh, yes, I was!
Simplify
Having Morten's interest boosted this project from a nice little side project that I'd probably just have posted on BGG to something more serious. By that time in November 2016, I had already worked a lot with Morten and David. I know how they think, and I have learned a lot about Automas.
Priority number one is that Automas must be as simple to handle as possible. Looking at my current version I found that although I liked how human-like Automa was in handling the buttons, fiddling with them each Automa turn (gaining, spending, counting) was too much work. Patchwork has quick turns, so the solo mode must be similarly quick to handle. Could I get away with Automa not caring about buttons all? How would that work? More importantly, how would scoring look like without buttons?
I found the answer by coincidence. I tend to do a lot of statistics when designing. I keep track of all kinds of numbers from my playtests. One day, when I looked through my spreadsheet, I realized that the number of buttons on patches Automa gained throughout a game was pretty close to my average score. Could I use that as Automa's score?
I created a deck of cards that gave Automa a virtual number of buttons to spend on her turn. This was less realistic, but easier to handle. You flip over a card that shows the number of buttons Automa can spend this turn. All available patches with a cost equal to or less than that are eligible for her to take. Give the patch to Automa. Discard the card. Done. By that time, the tiebreaker list was still fixed and unchanging.
Next, I got rid of Automa's quilt board. By that time, it tracked only the spaces needed for Automa to gain the 7x7 tile. My spreadsheet also included notes about the position of my time marker on the time board at the moment I claimed the 7x7 tile. I took an average, marked that space on the time board and tossed Automa's quilt board. When her time marker reached the marked space before I got the 7x7 tile, she claimed it.
This was the slimmest version of my Patchwork Automa. At the end of the game, Automa scored the number of buttons on her claimed patches (plus possibly the 7x7 tile). It was different than the regular scoring, but it allowed me to get rid of unnecessary and fiddly elements.
Gaming Automa (Part 1)
One of Morten's biggest fears is that players will find a way to game one of our Automas. We had several discussions in which he didn't want a player to have specific information of what Automa might do on her next turn. Most of the time he's right about that, but every now and then I argue that the specific information he wants to hide would also be open information in a real game. It's always an interesting fine line how realistic you want to be in that regard.
The fixed tiebreaker made Automa pretty predictable. Sure, you wouldn't know how many buttons Automa could "spend" on her turn before you drew the next card, but you could always tell that if Automa had a specific number of buttons, she would go for the available tile that's first in the tiebreaker list, so I removed the list from the rules and added a list to each card, mixing up the tiebreakers differently to make them unpredictable. A good side effect of this change was that I could adjust the tiebreaker list to the number of buttons Automa has for her turn. On cards with high numbers, the first tiebreakers tended towards "patch with most buttons" or "largest patch", while cards with low numbers focused on getting double turns (patch that costs least amount of time).
Tactical Variant (Part 1)
Having Automa's button numbers on cards added a level of unpredictability that I liked. Each Automa turn started with a card flip and a surprise. On the other hand, I realized that this didn't match the real game in which the number of your opponent's buttons is open information. More experienced players might miss this tactical element.
I made a version in which the button numbers were also shown on the back of the cards. You always knew Automa's buttons for her next turn. I tried it and found it boring. The entire element of surprise and tension was lost. We decided to postpone this for beta testing.
Scoring Ranges
The system worked pretty well, but after a dozen more playtests I realized that Automa's scoring didn't have enough variety. She always scored around 22 points, with a deviation of around +/- 7 points. A real player's scores can reach from -20 to +45 points or even beyond that.
I tried to introduce difficulty levels by changing the card distribution in the Automa deck. Three cards were the key here, one with 0 buttons, one with 4 buttons, and one with 10 buttons. In a regular game, you'd leave out the 4-button card; in an easy game the 10-button card; and in a hard game the 0-button card. This looked good on paper but only marginally changed Automa's scoring range.
I tried a lot of different card sets with different button distributions. I introduced different spaces on the time board where Automa gains the 7x7 tile. None of this worked well. We had a "too perfect" Automa with not enough variety in scoring.
Then Morten sent me a suggestion: Each time Automa's time token crossed an income button icon on the time board, she'd gain a number of buttons equal to the buttons shown on the patches she already gained. I didn't like it first because I was so happy that I could get rid of any kind of button income, and second because the number of buttons on gained patches is not under my control. It would be pure luck if this system worked.
On the other hand, Morten was right. We needed a random modifier to stretch Automa's scoring range and a trigger for when to apply it. I divided the overall Patchwork scoring range into four sectors from beginner to pro player, set a center number for each range, and calculated the +/- deviation needed to cover the full range. It was somewhere around +/- 7. I kept Morten's trigger suggestion (passing an income icon on the time track) and added a new value to each Automa card. This would give Automa somewhere between 0 and 5 buttons when she passed an income icon. The distribution of my numbers needed to be so swingy that it creates a random but still controllable element. I needed an overall bonus from anywhere between 0 and 15 within a game, which would comprise my +/-7 deviation.
If this worked well, I could even add four different values on each card to use for each of the four difficulty levels. When you play easy, use the first value; when you play hard the fourth value, etc.
It took a ton of iterations to find the right distribution on the cards, but in the end, it worked. Automa's score was less consistent now. For me, this was the first time in a design where something was too perfect and needed to be made artificially worse. Interesting experience...
There was a good side effect to the new system. When Automa had lots of buttons to spend, she most likely gained a valuable patch. In such cases I didn't want to give her additional buttons if she crossed an income icon, so I placed 0 button income on high button cards. On the low button cards where Automa was likely to gain a weak patch or even needed to pass, I added the high numbers. This was an unexpected lever to work with and added a fun little twist.
The downside was that we added a bit more fiddliness back to the system in dealing with button income for Automa and having to count them for scoring — and of course since we were now adding 0-15 buttons to the endgame scoring, the basic score needed to be lowered.
Difficulty Levels (Part 1)
Again, I consulted my spreadsheet with all kinds of numbers and statistics. There were multiple variables I could use for scoring:
• 7x7 tile
• Buttons gained
• Number of patches without buttons gained
• Number of patches with buttons gained
• Number of buttons on patches gained
I looked for combinations of those variables to make for a good scoring result. Once I found them, they looked somewhat like this:
Even better, I realized I could cover all difficulty levels just by adding different variables at the end of the game. This had two advantages. I'd need only one income icon on the cards, instead of one for each level, and the difficulty adjustment had no in-game mechanisms that you needed to remember while playing.
In addition, I introduced different spaces on the time board at which Automa would gain the 7x7 tile.
Beta Playtesting
In the meantime, Morten finished some of his other projects and had more time to join in. He had already contacted Lookout Games, and they showed interest. We knew this would not justify a new edition of the game or a standalone Automa expansion, but Lookout publishes its own magazine — "Neues vom Ausguck" — and talked about publishing our Automa in one of those issues. That was motivating.
We decided to start beta playtesting, and I'm glad Morten had enough attention on BGG to gather a playtester pool quickly. We did three waves of playtesting, starting with a small group of around ten players.
I was very happy with the initial feedback as Automa got good ratings right from the start, and it looked like I hadn't overlooked any major game-breaking factor. What a relief!
Difficulty Levels (Part 2)
After about eighty reported games, we had enough data to start tweaking the difficulty levels. Up until playtesting, I was more or less the only one who had played this excessively. All levels were more or less based on my personal skills and playing style.
We set target numbers for four levels and added the compulsory "brutal" level 5. Our targeted win ratios looked like this:
• Level 1: 80%
• Level 2: 50%
• Level 3: 25%
• Level 4: 10%
• Level 5: 1%
But first we had to deal with another problem: We kept track of how often players would gain the 7x7 tile. This tile feels pretty crucial in the game. Unless no one gets that tile, it's a 14-point swing. One player gains 7 points and also denies the other player the option to gain those points himself. Personally, I think this tile awards too many points. (My preference would be 3 or 4 points.) But as an Automa designer you should never change any original game concepts or rules. The problem was that the ratio in which players got the 7x7 tile did not match the win ratios of the games.
We needed to adjust the 7x7 win ratios to our desired difficulty level win ratios first before we tweaked scoring. I took our playtest data with the four positions of the 7x7 marker, added them to a time board, and filled the spaces in between with interpolated values.
Based on this, we set new positions on the time board.
Again, we needed to check our scoring variables: 7x7, buttons gained, empty patches gained, patches with buttons gained, and number of buttons on gained patches. Luckily, we found combinations that — at least on paper — would provide the win ratios for which we were aiming.
More Playtesting
In the meantime, Lookout had posted some hints about a Patchwork Automa on Facebook.
We started another wave of playtesting with the new variables, and they seemed to work well. Fifty more playtest results later with us collecting results and crunching numbers, we were finally there.
I cannot stress enough how valuable playtesting and playtesters are. Without them, you'll never be able to get 130+ games played with different approaches, skill levels, and tactics within a couple of weeks.
I'm very grateful to each and every player who contributed here. Without these players, this Automa wouldn't be as solid and balanced as it is now. Sounds like a cliché? Well, yes — but it's so true.
Tactical Variant (Part 2)
Still there was one open question we postponed as long as we could: Should the player know how many buttons Automa had for her next turn or not? We needed to bring this to the table again and rely on playtester feedback. We created two versions:
Version 1: We put the number of buttons on the back of the cards so that players could see them before the card is drawn. This came with another problem: Hardly any cards had the same button value. If a player memorized the cards based on their button value, they could derive the full card from the value shown on its back.
Version 2: Here we left the numbers on the front of the cards only, but on an Automa turn she would use the button number of the topmost card of the discard pile. In other words, the previously played (and discarded) card will show the button value for the next turn. The player wouldn't know which card will be drawn next and which tiebreakers might come, but they would know the number of buttons Automa would have. The huge downside here was that the button value was suddenly disconnected from the tiebreakers and button income value. All the interaction between these factors was lost.
In the end, we decided to use version 1. We changed the button values so that there are at least two of each in the deck. While players could still try to remember which cards have shown up, this would make it harder for them to memorize the deck.
By that time, we knew Lookout was going for a magazine publication, which allowed up to 24 cards. Since playtesters seemed to be almost evenly split about having the tactical variant over the non-tactical variant, we simply decided to add both.
A Few Cards
A few cards and a sheet of rules — this is what you get. Looks easy, doesn't it? It's been the first time I had the opportunity to go through such a meticulous development process of one of my own ideas. I'm glad it was "only" a "small" project. I learned a ton, and I'm even more happy that it's finally published, played, and appreciated. Lookout did a great job adapting the cards and rules to the original Patchwork style. Now you can enjoy it and play this fantastic game solo. Have fun quilting!
Thanks to Morten, David and everyone who helped.
Lines J. Hutter