Category: game design

My Excel skills have levelled up since I last wrote about balancing production line using player statistics. As a result I now have more informative charts to look at when analysing play sessions from build 1.32. My intentions with this balancing are to increase the long term playability and balance of the game. basically player retention is good after 1 day, good after 7 days, but starts to tail off before 28 days, implying that the game is good initially but loses its challenge after a while. it may also suggest a lack of content, which is surprising given what’s in the game, but will be naturally fixed over time as more is added (Pickup trucks, quality control, branding, breakdowns).

Looking at the following chart I can see that the amount of cash players have after 50,200,200,300…500 hours. I’m quite happy with this. clearly the amount climbs over time, but is not exorbitant for the median player. I’d like the player to have the odd million dollars in cash, but beyond 10 million makes things a bit easy. Hopefully some expensive upgrades for luxury cars in the late game will push that down slightly.

This second chart shows the intensity of AI competition, and is basically a measure of how well the player is doing, as perceived by the AI. I can see that I was absolutely right to do away with the 50-hour moratorium of AI competitors, as clearly some players race ahead and needed to have the AI rein them in. The clear problem here is that the competition value is trending rapidly up to 100%. I feel that this is a strong indicator that the maximum competitive level of the AI just is not competitive enough. In other words, the metrics by which the AI judges the player are not being bought under control by the methods available to the AI. This needs fixing.

This final graph shows the profitability (as percentage margin) of the players business over time. Its not unreasonable for this to be low, even at a loss during the start, as the player invests in equipment and ramps up production. Over time this is trending to slightly above zero, and my raw stats show an average value at 500 hours of 7.2%. This isn’t too bad, certainly believable in an industry like car production. I dont see that anything really needs to change is response to this graph.

So my conclusions from the currently available data is that the competition index metric is too meek, and that the player should face potentially more challenging AI at the top end, but at the bottom end, it should definitely continue to act as before, taking its foot off the metaphorical gas pedal of competition. The AI seems ok at not crushing the poor-performing player, but too weak to offer a decent challenge to the high-performing one.

Of course the important thing here is to work out what my ideal metrics are for improving the game. I’m assuming that people only continue to play games that they enjoy, and thus the hours played of the game should be a decent metric to show whether or not the game is getting more fun. Right now those stats look like this:

Which isn’t too shabby. I compared it with another one of my games and this isn’t too bad, especially considering the much shorter time its been out, and the fact that it is not content complete. Ideally you dont just make a game for those hardcore who put in 20+ hours but try to move everyone along that graph. I’d like to see the number of people playing 2 hours go up a lot more. I think if you don’t like a game you find out before then, so that’s a sign I’ve made something enjoyable. To that end, I need to ensure the game remains challenging in the long run, so tweaking these figures should hopefully nudge it in the right direction.

I feel I should do some actual marketing fluff here, so if you like the sound of the game and haven’t bought it, here is a link :D

I’ve talked about this issue in the design of customer AI in Production Line before.

In the last patch, I made some changes. here is the current system:

Each customer arrives at the showroom and looks at the cars on sale. That customer has a fixed ‘budget’ and have some leeway around that budget, from 20% less to 20% more (so a $20,000 customer checks out cars between $16,000 and $24,000, regardless what price range this puts them into). Every customer looks at every car and calculates a ‘score’ for that car…

They take into account the value of the car by comparing its estimated fair value to its actual value (basically they look at the markup you set). They then get a value from 1% to 100% saying how likely they are to buy that car. if the car is a different body style to the one they had originally wanted, they penalize that score.

The top five cars by this rating system are then looked at, and the player effectively rolls a percentage dice against each one to see if they will buy them. They may buy one of them, or not buy at all. The other four cars (or maybe all five) get given feedback by this customer on why they did not get bought, with the options being:

• Wrong Body Style (assuming thats true)
• Too Expensive ( failed the random die-roll)
• Missing features (The car was missing some essential features, and this had a 5% or more impact on the likelihood to buy.
• Bought an identical model (The customer bought exactly this model, but there was just more than one).

So like I say…thats the current system. It appears to have problems.

The most obvious problem is the customer budget. A top budget makes sense, but a bottom budget kind of does-not. If the customer wants a top feature sports car, and has a set budget of $200,000 and we are trying desperately to sell them for $100k, they should snap that up!. This is clearly nonsensical. What the customer should have is reasonable feature requests, not the minimum budget (which was being used as a proxy for this). The problem is, I need to do this sensibly, accurately, correctly and also fast, because some people have a LOT of cars on sale and a lot of customers. So how can I do this…

Right now I think the first thing I’m going to try is to remove the lower budget limit, but instead represent it as a quantity of features, that at a reasonable price, would be equivalent to that value. In other words, if My budget range is $80-120,000, I actually cap my buying at $120,000, but will consider any car that has $80,000-worth of features, regardless how far below $120k that car is priced.

Quick post before I stop work to eat!. I’ve been analyzing some play stats from sessions of the game to try and work out why some people are saying the game is impossible, and others are clearly generating a TON of cash. Some are obviously cheating the config files, but regardless of that I feel that there is definitely a problem with people amassing too much cash too early (in terms of providing a fun, balanced experience). Here is the cash balance of players at 50,100,200,300,400 and 500 in-game hours of play:

Its clear that there is a VAST range, and perhaps nobody should really be able to go beyond the 100million cash point so early, so I need to toughen up the cash in those first 50 hours. ironically the toughening up seems to overcompensate late in the game, as the trend is clearly to lose cash later, rather than increase it. yikes…

This second chart shows how strong the AI competition is during the game. I have made major screwups here, for example, the AI ignores you for the first 50 hours. BIG mistake, I need a lot more intelligence in deciding when to step in…

Lastly lets take a look at the players profit margin over time:

Clearly some cheating going on with those outliers, and some real disastrous, almost snapchat level loses for some people. Looks like that free fifty hours of no competition is letting people go nuts in the first 50 hours, and as I suspected from the above data, I then overcompensate and crush people with over-competition until their cash levels drop down.

I’m so glad I have this data, it will allow me to make a much better, much more balanced game.

The current system for customer simulation in Production Line is a bit too simple.

There are 4 bands of pricing, and multiple car body styles. Every customer, when created is assigned a price band and body style. They then show up at the showroom and see what is available. For each car that matches exactly their price range and body style, they rate that car based on their chance of buying it at the given price. There is a graphed slope based on the equation y = 1-(x^2) that gives the probability of buying a car, plotted against the relative premium that is being charged on it (or discount, obviously). This enables them to then pick a ‘best’ car (the one with the highest score).

That score is then the probability of the customer buying the car, which they then may or may not do based on a random number generator.

The positives to this system are:

• Overpriced cars take longer to sell, under-priced cars sell easier.
• There is a clear market segment for each body style.
• There is a clear market segment for each price band.

These are all good things. However, it has problems:

• As body style options increase, it seems unrealistic. Is there nobody who shops for an SUV who ends up with a sedan etc?
• Price bands are fixed. This means that there is no difference in pricing at the low or high end of the price band, as long as the markup is the same.
• The customer only really has a ‘chance’ of buying one car, the best one. Surely a wide range should result in a higher chance of a sale?
• The customer seemingly has no opinions on features. The presence or not of a sunroof, for example never swings (or loses) a sale, the player can always afford it, if its in their price-range.

So here I’m thinking of ways to improve on this.

Firstly I can change customers so they have a price point, and some fuzz. So rather than be looking at all cars from 0 -$15,000 (budget), they can have a price point of $14,000 +/- 20%, and thus evaluate any car from $11,200 to $16,800. This means we suddenly have an incentive to make some ‘low end’ budget cars (with some missing features) and some ‘high-end’ budget cars, with m,ore features, effectively catering to more tastes, and dissuading the player from just feature-cramming every model. On the downside, this also makes the list of designs maybe quite big… This would give me cool customer feedback about being too pricey, or lacking in features depending how things go.

Secondly I should have some fuzz around body styles. maybe a customer wants a sedan, but only applies a 20% negative modifier to a car of a different style. After all, if all the sedans are overpriced and crap but the sports car is surprisingly cheap and awesome, shouldn’t there be a chance of a sale?

Thirdly, maybe features should come into it. How about assigning a ‘must-have’ feature to each customer (thats common or universal at that price point), and they refuse to buy any car that does not have that feature? This would allow me to store some pretty fun customer feedback on the car ‘doesn’t have electric windows’ for example…

Fourthly, the wide range should come into it. Maybe the player should be able to consider up to five different cars (discounting ones with similar options), and run the random check against them all. This rewards the player for having a wide range, even within a single price band and body style.

That involves a bit of code, but as always with big sim games, the code is trivial (when you have coded a lot), its the design, the balance, the GUI and the making the code run super-fast (so having 100 customers how up and evaluate all this is not a frame-skipping event) that will take all the time. I’d like to know what people think, and if any of this sounds like the wrong direction before I start coding it. Obviously numbers given are easy for me to tweak, its just the pricing and sales mechanics I want to improve upon.