In a previous post I suggested that early in a relationship hetero women prefer men who show only slight interest. I argued the optimal early dating strategy is to show only as much interest as she does and cultivate a state of “Is he into me or does he just want to be friends”. This RCT is intended to test the theoretical issue, and advise future Tinder and Bumble policy specifically.
Method:
I will randomly select Tinder/Bumble matches to receive disinterested, low-energy messages and the rest will receive control messages. Matches will be split by flipping a coin. My swiping will continue as normal, as the algorithm is weird about it. The target sample size is 20.
The control group will receive my usual messages (Todd’s). They begin with a comment on a mutual interest or interesting fact about the person from their profile. If I see nothing interesting on their profile, I will introduce a topic I would like to talk about. Average message length is 3-5 lines, and double messages are common. Control group will occassionally receive emogis and exclamation points as an expression of interest. This is my regular texting style. After 6-7 messages I will invite the person to a coffee date. Ghosters will receive a second, in-character message on an unrelated topic after 48 hours.
I altered the treatment variable on advice from a colleague.
In the treatment case I only write messages that can be responded to in 5 seconds.
Never use exclamation points or emojis
Steer the conversation to something vaguely praiseworthy about the other person. Reward them for the statement. Try to repeat that a couple times. Only reward attributes that are plausibly praiseworthy, do not reward just similar attributes “that’s so cool, we both like Chinese food”
Respond to messages as quickly as possible but never respond when you lack time for a full conversation
Removed all standardization of response time
Analysis:
Control and treatment will be compared on the following four variables: number of first responses, number of total responses, number of coffeedates, and number of ghostings.
Edited because I changed the treatment on advice from a colleague.
Changelog
removed all discussion or response time, I just respond as quickly as possible when I can have a conversation (never during work hours)
big changes to treatment variable
I can’t reward people for religious statements because I am too athiest, other than that the rewarding is going well
Replacing control group with my previous messages to save time
But anyway, this little optimization is nothing compared to a deep toolbag of conversation skills you could employ. Feel free to DM me for more specific advice.
Update on day 3 - I altered the treatment variable on advice from a colleague. In the treatment case I only write messages that can be responded to in 5 seconds. Steer the conversation to something vaguely praiseworthy about the other person. Reward them for the statement. Try to repeat that a couple times
For the control group I just use my old tinder convos, that’s easier.
So far the results show massive increases in response rates. I’m stunned.
So the biggest change was to make wayyy simpler questions. I’m like… mad at myself for not noticing that earlier. But also like, if someone ghosts you have no idea why, so...
At least we know reductionism is a good philosophy. Occams razor: because it works.
Preregister a Tinder Randomized Control Tiral:
In a previous post I suggested that early in a relationship hetero women prefer men who show only slight interest. I argued the optimal early dating strategy is to show only as much interest as she does and cultivate a state of “Is he into me or does he just want to be friends”. This RCT is intended to test the theoretical issue, and advise future Tinder and Bumble policy specifically.
Method:
I will randomly select Tinder/Bumble matches to receive disinterested, low-energy messages and the rest will receive control messages. Matches will be split by flipping a coin. My swiping will continue as normal, as the algorithm is weird about it. The target sample size is 20.
The control group will receive my usual messages (Todd’s). They begin with a comment on a mutual interest or interesting fact about the person from their profile. If I see nothing interesting on their profile, I will introduce a topic I would like to talk about. Average message length is 3-5 lines, and double messages are common. Control group will occassionally receive emogis and exclamation points as an expression of interest. This is my regular texting style. After 6-7 messages I will invite the person to a coffee date. Ghosters will receive a second, in-character message on an unrelated topic after 48 hours.
I altered the treatment variable on advice from a colleague.
In the treatment case I only write messages that can be responded to in 5 seconds.
Never use exclamation points or emojis
Steer the conversation to something vaguely praiseworthy about the other person. Reward them for the statement. Try to repeat that a couple times. Only reward attributes that are plausibly praiseworthy, do not reward just similar attributes “that’s so cool, we both like Chinese food”
Respond to messages as quickly as possible but never respond when you lack time for a full conversation
Removed all standardization of response time
Analysis:
Control and treatment will be compared on the following four variables: number of first responses, number of total responses, number of coffeedates, and number of ghostings.
Edited because I changed the treatment on advice from a colleague.
Changelog
removed all discussion or response time, I just respond as quickly as possible when I can have a conversation (never during work hours)
big changes to treatment variable
I can’t reward people for religious statements because I am too athiest, other than that the rewarding is going well
Replacing control group with my previous messages to save time
I wrote a blog post on this topic—https://textgameformen.com/2014/04/08/gradually-figure-out-that-she-meets-your-high-standard/
But anyway, this little optimization is nothing compared to a deep toolbag of conversation skills you could employ. Feel free to DM me for more specific advice.
Update on day 3 - I altered the treatment variable on advice from a colleague. In the treatment case I only write messages that can be responded to in 5 seconds. Steer the conversation to something vaguely praiseworthy about the other person. Reward them for the statement. Try to repeat that a couple times
For the control group I just use my old tinder convos, that’s easier.
So far the results show massive increases in response rates. I’m stunned.
So the biggest change was to make wayyy simpler questions. I’m like… mad at myself for not noticing that earlier. But also like, if someone ghosts you have no idea why, so...
At least we know reductionism is a good philosophy. Occams razor: because it works.
Forecast- Conditional on the results being conclusive − 2:1 odds in favor of Aloof Alfie