So what we do at my house to randomly pick games that are 1v1 is:each opponent picks 1 expansion (where cornucopia/alchemy/guilds/promos count toward .5 of an expansion each); take the randomizer card deck; shuffle it; draw 10 card from the expansion (or 2 sets of 5 for the .5 expansions or to compensate when we pick a 1.5 sized expansion a 7/3 split between 1 whole expansion and a .5 expansion size [the math isn't perfect, I know]); decide on landscapes depending on what the game calls for individually and normally not randomize when we pick those at the end BUT if you were to create a game around a landscape card it would be 1:3 for events, 1:2 for landmarks, and 1:2 for projects where the equation means landscape required:random landscape available. I prefer a smaller sample size when choosing cards at random to base a game on because it's more of a challenge but maybe the landscape selecting ratio would fluctuate depending on how many players are playing.
If it is done outside of a 1v1 match and of course because there may be 4 players who want to choose only 2 of 3 landscapes the algorithm to randomize the landscape selection would have to work on a round robin turn order where said algorithm uses the last number of total games played on the server to decide what order of landscapes would be generated e.g., game #100 = event; #101 = landmark, #102 = project ... #108 = project, #109 = landmark, #110 = event, #111 = landmark, etc, etc, etc. (Not sure how even the distribution is in that model).
However, the upside to selecting a kingdom in this manner is that a player can choose which expansion he is comfortable with while also having an variety of cards at their disposal to build kingdoms with synergy/anti-synergy if the kingdom is to be selected in such a manner, or still be given the option to leave it to random chance (depending on what agreement has made between your company and the creator of Dominion; e.g. customers who don't subscribe to silver or gold memberships and must wait for random kingdoms with 5 or more expansions [not completely sure of the details as I have only skimmed over the FAQ]) Of course those details may not even matter because why allow someone who doesn't pay for subscription any arbitrary decision making regarding a custom kingdo.
AND ADD THE RECOMMENDED KINGDOM SETS THAT ARE IN THE BACK OF THE INSTRUCTION MANUALS OR FIND THEM ON WIKI.DOMINONSTRATEGY.COM IT CAN'T TAKE THAT MUCH MAN POWER OR TIME TO CODE PREDETERMINED KINGDOM SETS INTO THE KINGDOM SELECTOR, THIS BENEFITS NEWER PLAYERS AND WILL HELP GROW THIS BRAND AND ULTIMATELY MAINTAIN YOUR BOTTOM LINE SUBSCRIPTION SALES. HOW MANY PLAYERS HAVE NOT BOUGHT A SUBSCRIPTION TO DOMINION.GAMES BECAUSE THEY EITHER HATE/DON'T UNDERSTAND THE CARD MECHANIC'S SYNERGIES OR AREN'T ABLE TO BUILD A BASIC VILLAGE/SMITHY ENGINE BECAUSE THEY HAVE NEVER HEARD OF FIRST GAME. GUYS AT SHUFFLE IT LISTEN TO ME: FIRST GAME SHOULD BE THE FIRST GAME OF DOMINION THAT ANY PLAYER PLAYS, THOSE TWO WORDS HAVE A VERY LITERAL AND DIRECT MEANING.
I'm done, cheers.