The results posted by the OP that claim this were not independently checked or duplicated. Moreover, there is no proof that the posted server request represents the full communication between player and server for the game event.
I have checked and can confirm indeed the post by the game in free play and real play are different.
I recorded this XML message at Bet365, in real money mode, betting on black:
<placebet signature="xxx" site="bet365" session="xxx" stamp="xxx" game_instance="xxx" mode="REAL" game="HILOGAMBLER" initialnumber="0"><bet id="0" payout="null" repeatcount="1" stake="5" state="0"><betselections><betselection savedstake="0" selection="HIGHLOW_BLACK" id="0" /></betselections></bet></placebet>
And the response:
<placebet game_instance="xxx" playduration="28" session="xxx" stamp="xxx"><bet bettype="ACS" id="xxx" isqualifiedforjackpot="false" payout="9.75" repeatcount="0" repeatssofar="1" stake="5.0" state="PLC"><betselections><betselection id="0" price="1.95" selectedvalue="" selection="HIGHLOW_BLACK" stake="0.0"><betoutcomes><betoutcome haswon="1" price="1.95" value="2"/></betoutcomes></betselection></betselections></bet><gamestate isover="0" isspecial="0" iswaitingstake="0" iswon="1" manuallypaidout="0" midbalance="xxx"><result number0="2"/></gamestate><account balance="xxx" country="GB" displaybalance="xxx" firstname="" language="1" lastname="" nonwithdrawablebalance="NaN" playabilityrolloverpercent="0.0" title="" username="xxx"><preferences mus="0" snd="0"/></account><notifications/></placebet>
And at Betfred in free play with the same bet:
<funplayrequests signature="xxx" site="BetFred" mode="FUN" game="HILOGAMBLER"><funplayrequest id="0" allowduplicates="true" count="1" rangehigh="12" rangelow="1" method="generaterandomnumbers" /></funplayrequests>
And the reply:
<funplayrequests><randomnumbers id="0">10</randomnumbers></funplayrequests>
My observations are:
* At Bet365 the game pays out 1.95, which is consistent with a fair 97.5% RTP game, but I haven't tested it for fairness.
* It is necessary for a real play game to send the bet at the time of requesting the result, so the message request in real mode is not inherently suspicious
* It is not unusual for a free play game to operate 'offline', for speed and to conserve server resources, albeit that technically no server request is required at all, since the random numbers can be generated client-side
* It is indeed correct, that if the real play game is adaptive to player's bet, that the free play game cannot be, since I can confirm that no message is sent to the server other than these that could contain the player's bet.