All posts


Game development








Hello! I'm Tom. I designed a game called Gunpoint, about rewiring things and punching people, and now I'm working on a new one called Heat Signature, about sneaking aboard randomly generated spaceships. Here's some more info on all the games I've worked on, here's the podcast I do, here are the videos I make on YouTube, here are some of the articles I wrote for PC Gamer, and here are two short stories I wrote for the Machine of Death collections.


By me. Uses Adaptive Images by Matt Wilcox.

Heat Signature’s Launch, And First Player Legend

A Leftfield Solution To An XCOM Disaster

Rewarding Creative Play Styles In Hitman

Postcards From Far Cry Primal

Solving XCOM’s Snowball Problem

Kill Zone And Bladestorm

An Idea For More Flexible Indie Game Awards

Teaching Heat Signature’s Ship Generator To Think In Sectors

What Works And Why: Multiple Routes In Deus Ex

Natural Numbers In Game Design

Naming Drugs Honestly In Big Pharma

Writing vs Programming

Let Me Show You How To Make A Game

New Heat Signature Video: Galaxies, Suction And Wrench-Throwing

What Works And Why: Nonlinear Storytelling In Her Story

My Idea For An ‘Unconventional Weapon’ Game

From Gunpoint To Heat Signature: A Narrative Journey

The Cost Of Simplifying Conversations In Videogames

What Works And Why: Invisible Inc

Our Super Game Jam Episode Is Out

What Works And Why: Sauron’s Army

Showing Heat Signature At Fantastic Arcade And EGX

What I’m Working On And What I’ve Done

The Formula For An Episode Of Murder, She Wrote

Heat Signature Needs An Artist And A Composer

Improving Heat Signature’s Randomly Generated Ships, Inside And Out

Gunpoint Patch: New Engine, Steam Workshop, And More

Distance: A Visual Short Story For The Space Cowboy Game Jam

Raising An Army Of Flying Dogs In The Magic Circle

Floating Point Is Out! And Free! On Steam! Watch A Trailer!

Drawing With Gravity In Floating Point

What’s Your Fault?

The Randomised Tactical Elegance Of Hoplite

Here I Am Being Interviewed By Steve Gaynor For Tone Control

Heat Signature: A Game About Sneaking Aboard Randomly Generated Spaceships

The Grappling Hook Game, Dev Log 6: The Accomplice

A Story Of Heroism In Alien Swarm

One Desperate Battle In FTL

To Hell And Back In Spelunky

Games Vs Story 2

Gunpoint Development Breakdown

Five Things I Learned About Game Criticism In Nine Years At PC Gamer

My Short Story For The Second Machine Of Death Collection

Not Being An Asshole In An Argument

Playing Skyrim With Nothing But Illusion

How Mainstream Games Butchered Themselves, And Why It’s My Fault

A Short Script For An Animated 60s Heist Movie

The Magical Logic Of Dark Messiah’s Boot

Arguing On The Internet

Shopstorm, A Spelunky Story

Why Are Stealth Games Cool?

E3’s Violence Overload, Versus Gaming’s Usual Violence Overload

The Suspicious Developments manifesto

GDC Talk: How To Explain Your Game To An Asshole

Listening To Your Sound Effects For Gunpoint

Understanding Your Brain

What Makes Games Good

A Story Of Plane Seats And Class

Deckard: Blade Runner, Moron

Avoiding Suspicion At The US Embassy

An Idea For A Better Open World Game

A Different Way To Level Up

How I Would Have Ended BioShock

My Script For A Team Fortress 2 Short About The Spy

Team Fortress 2 Unlockable Weapon Ideas

Don’t Make Me Play Football Manager

EVE’s Assassins And The Kill That Shocked A Galaxy

My Galactic Civilizations 2 War Diary

I Played Through Episode Two Holding A Goddamn Gnome

My Short Story For The Machine Of Death Collection

Blood Money And Sex

A Woman’s Life In Search Queries

First Night, Second Life

SWAT 4: The Movie Script

Analysing Team Fortress 2

Games And Graphs, Together At Last!

One of my only criticisms of Team Fortress 2 was that the Medic isn’t as fun to play as the other classes – a particular shame when he’s so critical to success. Some people objected to this, because they really enjoy the Medic, so I’d just like to make sure these people don’t miss the Team Fortress 2 stats: cold, hard evidence that I am right and they are wrong. The Medic, for anyone not motivated to click the link, is the least-played class by a head.


Obviously the THIRTY-SIX achievements they’re adding just for Medics is an attempt to redress that imbalance, but I wonder if that’s the way to fix this.


You never had to bribe people to play Medic in Battlefield 2, partly because they were effective combatants, but I think mostly because the medicking part of their job was extraordinary fun. When I’m holding my healing ray on a Heavy while he mows people down in TF2, I feel like I’m serving him. When I sprint through thwacking gunfire and dive defibrillators-first onto the unconscious body of my squad leader in BF2, I feel like I’m saving him.


What’s surprising about those stats- well, okay, there are lots of surprising things about those stats.

1. The first surprising thing is the very first fact: the Scout is the most-played class? He’s the only class which, for the majority of any given round, is almost entirely useless. The second the enemy have a single sentry up in any sensible location, he has no way of getting to their objective and is too weak to effectively defend his own. It must just be that, like me, a lot of people always play him for their first life on 2fort, well and granary. But when I do, can I persuade the rest of my team to get a decent Scout rush going? Can I testicles.

2. Speaking of those three maps – the three perfectly symmetrical ones – here’s the most remarkable stat of the lot: the Blue team is almost twice as likely to win on any of these. Even 2fort. These are maps in which each team’s base is a mirror of their enemy’s, and the game’s teams have no inherent differences. If you played in black and white, you wouldn’t even be able to tell which team you were playing on.

I can think of only two explanations for this, and the first one is stupid. Perhaps the Red team are just slightly easier to see? This would be a perfectly reasonable theory in a game with large maps or camouflaged players, but Team Fortress 2 is depicted with unprecedented clarity. It’s the one game in which you can always spot enemies and even tell which class they are, at any range. Perhaps Snipers, through the smallest and darkest of windows, sometimes go unnoticed for a moment, but you’d think Blue Snipers would stand out more strikingly against the warm wood buildings of well, and that’s the map with the the strongest pro-Blue bias of all.


The other possibility is that for whatever reason, better players pick Blue. It’s not often you get to choose your team, since one usually outnumbers the other when you join, but the times when you do could account for this difference. It would have to be an overwhelming trend, to show through the auto-balance and playercount restrictions, but it’s possible. I pick Blue when I can – maybe I’m just that good.

The real answer is probably the counterpart to this: it seems possible that new or inexperienced players might automatically pick Red, since it’s first on the team-choice menu.

3. My other criticism of TF2 was that hydro didn’t quite work. It’s the map that changes shape every round, in complicated ways, in order to keep it fresh for years. As far as I’d played when I reviewed it, this just seemed to keep it confusing for years, but I said I was prepared to bear with Valve’s experiment to see how it played out.

Three months later, I have a more conclusive answer: sucks! hydro is awful. But looking at the stats, Valve must be delighted: apparently hydro sees the longest rounds of any map, and never results in a stalemate. That’s funny, because around fifty percent of all stalemates I’ve ever had and my ten shortest rounds have all been on that very map.


The problem is that they don’t count the fight over two control points – before the map reconfigulates – as a round. They count the entire, tedious push through each arbitrary mess of blocked-off routes towards the enemy’s final base – at least four separate games – as one round. This cleverly conceals the two main ways in which hydro sucks: if one team is even slightly better than they other, they utterly storm the enemy control point in a matter of seconds, and no-one has any hope of mounting a comeback or even having an influence on the battle. And if the teams are even in skill, every damn game ends with two nests of Sentry Guns sitting vigilantly at their own bases, waiting for the Sudden Death timer to run out.

4. As you can see from the ugly grey lumps in that graph, Stalemates are all too common on the maps where they can occur. Amazingly the solution to this is so simple the community have already implimented it in places: a fantastic server-side mod causes everyone to spawn as the same class when entering Sudden Death, and restricts them to melee. When I last played on such a server, this meant twenty-four Heavy Weapons guys punching each other to death, but pretty much any class is as funny.


It completely transforms the dark, paranoid, defensive atmosphere of Sudden Death into a glorious burst of humour and madness at the end of the round. Instead of saying “You’ve failed to complete a game of Team Fortress 2, now you must play Counter-Strike until everyone gets bored and leaves or the game tells everyone they suck”, it says “Eh, you guys are about as good as each other. Fist fight! Woo!” Then it spins around with its arms out until it falls over from the giddiness. In other words, it’s silly and friendly and hilarious in just the way TF2 is everywhere else. You actually come away from it feeling almost like friends, instead of hating the enemy team’s stupid camping guts and your own team’s stupid non-Medic faces.


But if I were Valve, I wouldn’t be working on any of these issues yet. In fact, I’d be doing absolutely nothing to the game until I’d come up with the perfect auto-balancing/team-reshuffling algorithm. I think they ended up maximising almost every other factor that positively contributes to the percentage of time you spend enjoying a multiplayer game, but left alone the biggest one: engineering a fair fight.

If it were up to me, no-one would get to pick a team. Everyone’s auto-assigned according to their skill level, keeping Friends together and players who prefer the same class apart, in that order of priority. After every round, the highest-scoring player from the winning team, along with the third best, fifth best, seventh, etc, are switched with the second best from the losing team, and the fourth, and sixth, respectively. In other words, maximum rejiggling with a slight bias towards the losing team, giving the best players a challenge and the worst players a break.

The reason you couldn’t do most of this stuff in older games, like the original Team Fortress, was that the game simply didn’t have access to that sort of information about players. Steam now has all this and more, and if they’re only using that for playtesting, they’re missing the real value of this kind of data. They’ve got everything they need here to rig a multiplayer game to be fun every time, and that could be a hell of a thing.


Jack: hydro is the best map. i do not understand why everyone hates it.

roBurky: Hydro is awful. But it's nothing to do with the changing terrain.

It's the mechanics of it. It's because it is just two capture points, with both teams attacking and defending. Because there's only one point to take, it's quite possible to win a map segment within a few seconds of the start of the game, leaving both teams bemused.

But if that doesn't happen, and both teams manage to dedicate some defense, then taking that control point becomes a horrible chore because it is so close to the enemy's spawn point. There's no attacker's advantage like there is on the similar last points of well or granary. And unlike the two-point dustbowl mini-rounds, both teams have to split their attention between attack and defense, making it harder to maintain a momentum when you do manage to break through a defensive line.

So most mini-rounds go to sudden death. And almost all sudden death rounds end in Another. Fucking. Stalemate.

Tom Francis: The changing terrain doesn't help: the ten seconds you spend working out where the enemies can actually come from are all it takes for a Scout rush to get to your cap. Engies, the usual counter to rushes, are virtually useless without perfect situational awareness from second zero.

It also doesn't help that there are GIANT FLASHING ARROWS pointing to routes blocked off by LOCKED METAL BARRIERS, as if to say, "Hey guys, wouldn't it be nice if you could get through here?"

The_B: The arrows annoy me immensely - especially when they're pointing to barriers that AREN'T EVEN OPEN.

Although I think dynamic maps could work better eventually, Valve have a bit to learn yet though.

It's wierd though how some things can be attributed to the simplest of things, like you say perhaps more new players pick red because it's the first choice, but conversely I think that perhaps more people pick red first as a team choice, while the people who are actually playing in a sportsman like manner - and as elitist as this sounds, are probably the better players - choose the random door, and thus get put onto the blue team. Or as someone has said to me, even simpler: the blue door is the closest to the centre, so is the least amount of mouse movement used to get to it, and clicked first by people who just want to play.

grey_painter: I quite like Hydro, I think I've played it enough to have seen most of the arrangements of maps so I usually know where to go once I see the two entry points to the start area. I like having a few oft forgotten routes into each base too to exploit.

As for the blue/red split, I'm not convinced it is that statistically relevant. Then again I never did well at stats.

Man Raised By Puffins: With regards to the Medic stats, it's probably worth bearing in mind that as a pre-requisite to play as one you need a decent pool of players on your team to heal (or failing that at least a lone Heavy). If a server only has a few people on each team, the chances are that it isn't worth having a Medic on the team. Well, unless if said Medic is a crackshot with the syringe gun anyway.

Without seeing a more detailed breakdown though, it's difficult to tell whether I'm talking out of my arse here or not.

ImperialCreed: Am I the only one who doesn't have a favourite class? Most folk I've played with seem to stick to one class for a whole game, unless our team happens to be desperately in need of medic or a spy to force a breakthrough.

As an Engineer I love putting up and taking care of my turret, but I'm also fond of sniping Heavies from afar. And being a Scout becase they can annoy people so much.

I found Hydro to be an okay map, though it's the one I've spent least time on. It's Dustbowl I totally suck at.

JakethePirate: Well, I think Medic is fun and it is my most played class. My second most played class is Pyro, so maybe it's just me.

Melee heavy only sudden death is one of mans greatest accomplishments.

redtoade: When you join a server with only a handful of people playing, isn't it typical to have a bunch of scouts and no medics? That's the result of big holes in the defenses and no one to heal. Someone always goes scout and tries to cap at least once before the server fills up.

Engineers are always played, because bar none they are the most necessary class in the game. Not the most important, just that, no matter what the game conditions, it is required to have engineers playing in order to keep the game from becoming a blow out. They establish the defensive lines. They slow down the enemy's offensive, remove the advantage that the scout had during the game's first few minutes, and create the choke points. If your team becomes short handed, defense calls for engineer. If your team has the other pinned down in the respawns, someone goes engineer to setup a turret outside offensively.

So sure, scout and engineer plays all the time. Makes sense to me.

Soldier and demoman have rocket/grenade jump. This opens the map up so that they can go places the medic, heavy and pyro can not. I can understand why they'd be played more often.

That's why I love pyro. He has enough speed to cut off a scout if you know where he's going. Enough health to take a few shots and close the distance on an aware enemy. And a weapon that can make an entire squad of charging opponents stop and back up thus blowing their momentum. Ah the sneaky sneaky pyro.

Cptn.Average: Speaking from a bit of hindsight (it being a year since Toms post went up) one thing that strikes me is how the scout is now the least played class and coincidentally how hydro doesn't get a lot of play anymore. A link maybe? I know from my own experience that scouts have a lot of room to manoeuvre on Hydro (for the most part anyway). Most other maps just aren't as favourable for the scout, especially if the enemy gets time to prepare.

But for the record I love hydro about as much as dustbowl, which makes me sad to see it being underplayed =( Maybe if the cap points were more/less exposed people could enjoy it a bit more.